AI-based intelligent operation method and system for internet of vehicles commodities

Abstract

The application discloses an AI-based Internet of Vehicles commodity intelligent operation method and system, collects commodity operation data and competitive product market data, generates a pre-adjusted price trigger identifier through price rhythm coupling analysis, detects inventory backlog rate and establishes an adjusted price monitoring rule table in combination with a competitive product benchmark set, identifies an adjusted price gap category based on a favorites retention behavior, performs weight allocation on the adjusted price monitoring rule table to generate an operation control rule set, and carries out competitive product follow-up response speed over-limit checking to determine an intervention threshold node; according to the intervention threshold node, an adjusted price behavior sequence is activated, a recommended gap coefficient is generated by extracting a vehicle owner click non-purchase feature, a complementary product association recommendation trend monitoring is performed to generate a dynamic sales compliance degree, an intervention recurrence period is analyzed based on the dynamic sales compliance degree, a search word and a transaction category deviation rate are calibrated to generate a category operation portrait, and finally, an intelligent operation evaluation result is output by comparing the category operation portrait with the commodity operation data, thereby improving the precision and timeliness of the Internet of Vehicles commodity adjusted price decision.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation technology, and in particular to an AI-based intelligent operation method and system for connected vehicle products. Background Technology

[0002] Connected vehicle platforms aggregate a vast amount of transaction data and user behavior data related to automotive-related products. However, platform merchants often face the challenge of information lag when formulating pricing strategies. They lack systematic means to track competitors' price adjustments, making it difficult for merchants to determine the degree of deviation between their current pricing and the overall market level. This is especially true when similar products maintain stable prices in the market for an extended period; the underlying signals for price adjustments inherent in this stability are often overlooked, leading to repeated missed opportunities for proactive price adjustments.

[0003] Meanwhile, the connected vehicle platform offers a wide variety of products, with significant differences in sales rhythms, user decision-making cycles, and complementary consumption structures across different categories. Relying on manual experience for differentiated operations across categories is unsustainable. Data on user behavior—such as abandoning purchases due to price or category mismatch—is not being effectively utilized. The structural mismatch between search term traffic and actual transactions has long been unquantified and unidentified, resulting in a lack of objective basis for allocating operational resources across different categories. Consequently, overall operational efficiency cannot be continuously optimized through a traceable evaluation system. Summary of the Invention

[0004] This invention discloses an AI-based intelligent operation method and system for connected vehicle products. It aims to establish a dynamic price adjustment monitoring mechanism by jointly analyzing the price rhythm of competitors and the status of inventory backlog, identify operational gaps by combining user collection retention and click conversion behavior, achieve synergistic driving of price adjustment strategies and complementary product recommendations, and form a sustainable optimization category operation evaluation system through search term deviation rate calibration and multi-dimensional comparison.

[0005] The first aspect of this invention proposes an AI-based intelligent operation method for connected vehicle products, comprising the following steps: Collect product operation data and competitor market data, and perform price rhythm coupling analysis on the product operation data and competitor market data to generate a pre-adjustment trigger identifier; For the product operation data, perform inventory backlog rate detection to generate backlog rate distribution, obtain the competitor benchmark set through the competitor market data, and establish a price adjustment monitoring rule table by adapting the backlog rate distribution to the competitor benchmark set according to the product category sales cycle. Based on the pre-adjustment trigger identifier, the weight allocation of the price adjustment monitoring rule table is used to generate an operational control rule set. The operation control rule set is then used to conduct a competitor price-following response speed exceeding limit verification to determine the intervention threshold node. Based on the intervention threshold node, the operation control rule set is threshold-mapped to activate and output a price adjustment behavior sequence. The car owner click-but-not-purchase feature is extracted from the price adjustment behavior sequence to generate a recommendation gap coefficient. Based on the recommendation gap coefficient, complementary product association recommendation trend monitoring is performed on the price adjustment behavior sequence to generate sales compliance. Based on the sales compliance and the price adjustment behavior sequence, the intervention recurrence cycle is analyzed to generate an intervention recurrence cycle. Based on the intervention recurrence cycle and the operation control rule set, the deviation rate between search terms and transaction categories is calibrated to generate a category operation profile. The category operation profile is compared with the product operation data to generate an intelligent operation evaluation result.

[0006] A second aspect of this invention proposes an AI-based intelligent operation system for connected vehicle goods, comprising: The data acquisition module is used to collect product operation data and competitor market data, and to perform price rhythm coupling analysis on the product operation data and competitor market data to generate a pre-adjustment trigger identifier; The inventory detection module is used to perform inventory backlog rate detection on the product operation data to generate a backlog rate distribution, obtain a competitor benchmark set through the competitor market data, and establish a price adjustment monitoring rule table by adapting the backlog rate distribution to the competitor benchmark set according to the product category sales cycle. The rule allocation module is used to generate an operation control rule set by weighting the price adjustment monitoring rule table based on the pre-price adjustment trigger identifier, and to conduct a competitor price-following response speed exceeding limit verification to determine the intervention threshold node for the operation control rule set; The recommendation monitoring module is used to perform threshold mapping on the operation control rule set based on the intervention threshold node to activate and output the price adjustment behavior sequence, extract the car owner click-but-not-purchase feature for the price adjustment behavior sequence to generate a recommendation gap coefficient, and perform complementary product association recommendation trend monitoring on the price adjustment behavior sequence based on the recommendation gap coefficient to generate sales compliance. The results output module is used to perform intervention recurrence cycle analysis based on the sales compliance and the price adjustment behavior sequence to generate an intervention recurrence cycle, perform search term and transaction category deviation rate calibration based on the intervention recurrence cycle and the operation control rule set to generate a category operation profile, and compare the category operation profile with the product operation data to generate an intelligent operation evaluation result.

[0007] The beneficial effects of this invention are reflected in the following points: First, addressing the problem of lacking competitor rhythm awareness in pricing decisions for connected vehicle products, this invention extracts the rigid cyclical characteristics of competitors' historical prices and matches them with the price trend of the product itself. It identifies the cumulative deviation of the product against the backdrop of long-term price stagnation in competitors, and, combined with multi-category inventory backlog rate detection and new / old model substitution correlation correction, establishes a category-differentiated price adjustment monitoring rule system. This system expands static price comparison into dynamic perception of the rhythmic evolution of competitors, shifting pricing adjustment decisions from experience-driven to data-driven, especially for categories where competitors' rhythmic wait-and-see behavior is concentrated, demonstrating strong early identification capabilities. Second, the combined introduction of analysis of the distribution of time spent in favorites and verification of competitors' price-following response speed exceeding limits allows the identification of price adjustment gap categories to simultaneously rely on two dimensions: user price resistance and competitor behavioral threats. After implementing targeted weight allocation on the operational control rule set, the timing of price adjustments can achieve a balance between proactive intervention demand and the risk of competitors quickly imitating, solving the structural defect of ignoring the competitor's response rhythm when simply relying on inventory pressure to trigger price adjustments. Finally, the conversion gap quantification of car owner clicks without purchase behavior and the sales performance monitoring of complementary product association recommendations constitute a complete closed-loop verification link for recommendations, avoiding the problem of long-term lack of feedback on the effectiveness of recommendation strategies. On this basis, combined with the calibration of the deviation rate between search terms and transaction categories and the analysis of intervention reproducibility cycle, a multi-dimensional category operation profile is generated and compared with the actual operation data at different levels, outputting traceable intelligent operation evaluation results, providing an objective basis for the continuous optimization and allocation of operational resources among different categories. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating the AI-based intelligent operation method for connected vehicle products according to the present invention.

[0009] Figure 2 This is a structural block diagram of the AI-based intelligent operation system for connected vehicle goods according to the present invention. Detailed Implementation

[0010] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0011] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0012] The technical solutions of the embodiments of this application will be described below.

[0013] like Figure 1 As shown, this embodiment of the invention provides an AI-based intelligent operation method for connected vehicle products, including the following steps S11-S15: Step S11: Collect product operation data and competitor market data, and perform price rhythm coupling analysis on the product operation data and competitor market data to generate a pre-adjustment trigger identifier.

[0014] Specifically, product operation data and competitor market data are collected. The platform's transaction system is the direct source of product operation data. Historical transaction prices, inventory change records, page clicks and favorites, and promotional activity nodes for each product category are extracted in real time and entered into the data layer. Each data point is stamped with a time stamp accuracy at the hour level to ensure that the product operation data has accurate time positioning capabilities in subsequent price rhythm analysis. Competitor market data is periodically crawled from major competing platforms through the market monitoring module. The crawled content includes the current selling price, historical price adjustment records, and promotional status markings of competitors. The crawling cycle is consistent with the collection frequency of product operation data to ensure that the two types of data can be aligned and compared on the timeline. After collection, both product operation data and competitor market data undergo a unified data cleaning process. For sudden price increases or decreases on a single day—including records where prices deviate from the normal range due to promotional activities—the average of adjacent time points is used as a substitute with additional correction annotations to avoid promotional noise interfering with subsequent rhythm extraction. Records of anomalies triggered by competing products in the same category at corresponding time points in the competitor market data are subject to the same substitution rule. Aligning the cleaning standards of both types of data ensures that the comparability of subsequent matching is not affected by unilateral noise. Records with missing timestamps are not included in the price rhythm coupling analysis. The time density of collection behavior records in product operation data often shows a concentrated increase before and after holiday promotions. The collection module sets an independent time window for collecting behavior records, with a window length of 7 days, to ensure that short-term behavioral clusters before and after holiday promotions are not diluted by the monthly average. The window statistical results for the period of concentrated collecting behavior are included in the price rhythm coupling analysis process along with the overall product operation data.

[0015] In some embodiments, the step of generating a pre-adjustment trigger identifier by performing price rhythm coupling analysis on the product operation data and the competitor market data includes: extracting historical price sequences of competitors based on the competitor market data to construct a rigid rhythm distribution; performing price rigidity cycle alignment matching between the rigid rhythm distribution and the product operation data to generate an alignment association table; identifying price rigidity intervals where competitors have not lowered their prices for a long time through the alignment association table to generate an offset feature set; and generating a pre-adjustment trigger identifier based on the cumulative trigger strength of the offset feature set.

[0016] Based on competitor market data, historical price sequences of competitors are extracted to construct a rigid rhythm distribution. The construction of the rigid rhythm distribution starts from the temporal structure of the historical price adjustment records of each competitor in the competitor market data. After expanding the historical price adjustment records of each competitor in the competitor market data by product number and timestamp, it constitutes the historical price sequence of the corresponding product category. The continuous period when the price difference between adjacent nodes in the historical price sequence of the competitor product is zero constitutes a rigid descriptive unit. The three attributes of start node, end node and duration form the complete descriptive entry of the segment. After summarizing all the rigid descriptive entries of each competitor by product category, the frequency of occurrence and duration distribution of rigid segments in each category within the historical period are statistically analyzed. Categories with high frequency and concentrated duration exhibit strong rhythmic characteristics in the rigid rhythm distribution. The price observation behavior of these categories is periodically predictable, forming distinct periodic peaks in the rigid rhythm distribution, indicating a stable price stagnation pattern. This pattern will be used as an important reference for judging the timing of price adjustments in the subsequent matching phase. Categories with a multi-peaked duration distribution suggest that competitors have multiple overlapping observation rhythms before and after different seasons or promotional periods. The time periods corresponding to each peak independently trigger window sliding during the matching phase, avoiding a single peak from obscuring the effective signals of other rhythmic segments. Categories with frequent competitor price adjustments correspond to fewer rigid segments and weaker rhythmic characteristics, clearly distinguishing them from strong rhythmic categories in the rigid rhythm distribution. These two categories receive differentiated rhythmic weights during the matching phase. The rigid frequency distribution and duration distribution of all categories together constitute the rigid rhythm distribution, and the category coverage of the rigid rhythm distribution is consistent with the number of categories involved in the competitor market data.

[0017] A matching table is generated by aligning the price oscillation cycle of the rigid rhythm distribution with the product operation data. The matching is based on the oscillation cycle characteristics of each category within the rigid rhythm distribution. The product operation data provides the product's price time series as the matching object. The product's price time series in the product operation data is slid along the time axis to the oscillation segments identified by the rigid rhythm distribution, calculating the price fluctuation range of the product within the start and end range of each oscillation segment. When the fluctuation range is close to zero, it is determined that the product and its competitors are price-synchronized within that oscillation segment. Successfully matched competitor oscillation segments and the corresponding time periods of the product are recorded as a matching entry. Even when the product's price fluctuates slightly but does not undergo a directional adjustment, a price synchronization is also considered if the price range within the oscillation segment is less than 0.3 times the historical daily average fluctuation range of the category, thus avoiding missed matching entries due to normal price fluctuations. The density of matching entries reflects the degree of correlation between the product's pricing and the competitor's rhythm. Categories with high frequency of competitor price freezes and consistent synchronization of the product have a large number of matching entries, while categories with frequent competitor price adjustments and few freezes have fewer matching entries. These two types of categories are clearly distinguished in the matching association table. Matching entries for each category are arranged in chronological order by category number and freeze time to form the matching association table. The category coverage of the matching association table is consistent with the distribution of freeze rhythms and the intersection of categories on both sides of the product operation data. Categories with dense synchronization entries in the matching association table correspond to more effective offset feature entries during the offset feature set generation stage.

[0018] An offset feature set is generated by identifying long-term price stagnation intervals of competitors using a positional correlation table. Each entry in the positional correlation table is filtered based on the duration of the competitor's stagnation segment. The threshold for determining a long-term price stagnation interval is set at 1.5 times the historical average of the category. Competitor segments corresponding to entry in the positional correlation table that exceed this duration are entered into the candidate offset entry extraction process. Records in the candidate offset entries where the product's price also did not actively decrease during the corresponding period are considered valid offsets. The category of a valid offset, the corresponding time point, and the offset magnitude together constitute an offset feature entry. The offset magnitude is the normalized value of the absolute difference between the product's current price and the historical average price of the category. The standard deviation of the category's historical prices is used as the normalization benchmark for offset magnitude normalization to ensure cross-category comparability of the offset degree of categories with different price levels. High-priced categories with large absolute price differences will not have artificially inflated rankings in the offset feature set due to differences in units. When a product category exhibits effective offsets within multiple consecutive stagnant competitor price intervals, the magnitudes of each offset are sequentially superimposed, resulting in a continuously increasing cumulative offset. This reflects that the product category has not intervened to adjust prices despite the prolonged stagnation of competitor prices, and the unutilized price adjustment space accumulates as the number of stagnant intervals increases. Products with only a single, short-term effective offset have relatively small cumulative offsets, creating a significant gap compared to products with multiple consecutive offsets in the offset feature set. This gap directly affects the normalized ranking of subsequent trigger strength. The category coverage of the offset feature set is consistent with the set of categories in the alignment table where effective alignment entries are not empty. Categories with the highest cumulative offset magnitude in the offset feature set, ranking in the top 20% of all categories, exhibit the most prominent cumulative offset and correspond to higher normalized values ​​in the cumulative trigger strength calculation phase.

[0019] Pre-price adjustment trigger identifiers are generated based on the cumulative trigger strength of the offset feature set. The cumulative offset of each category in the offset feature set is normalized and compressed to the range of 0 to 1. The normalization results for each category in the offset feature set form a cumulative trigger strength value that can be compared horizontally. When the trigger strength is higher than 0.6, the trigger condition is considered met, and a pre-price adjustment trigger identifier entry is generated for the corresponding category. Categories with a trigger strength lower than 0.6 have insufficient current offset accumulation, and no identifier generation is triggered. A pre-price adjustment trigger identifier entry consists of three elements: category number, trigger time point, and cumulative trigger strength. A higher trigger strength indicates a higher urgency for proactive price adjustments in that category. Categories with a trigger strength of 0.8 or higher typically experience continuous offset accumulation within multiple competitor stagnation ranges without price adjustments, and the current price adjustment pressure has reached a high intensity state. These categories are recorded as high-intensity entries in the pre-price adjustment trigger identifier. Categories with a trigger strength between 0.6 and 0.8 have offset accumulation that has trigger basis but has not yet reached the most urgent level, and are recorded as medium-intensity entries. Categories with a trigger strength below 0.6 are still in the observable stage, and no corresponding entries are generated in the pre-price adjustment trigger identifier. The division of the three intensity ranges ensures that the pre-price adjustment trigger identifier forms an effective gradient distinction for categories with different accumulation levels. All category entries that meet the triggering conditions are arranged in descending order of cumulative trigger strength to form a complete pre-price adjustment trigger identifier. The category coverage of the pre-price adjustment trigger identifier comes from the subset of categories with a trigger strength higher than 0.6 in the offset feature set. When the number of trigger entries in the same natural month exceeds 30% of the total number of categories in the pre-price adjustment trigger identifier, it indicates that the current market as a whole is in the stage of concentrated price rhythm convergence, and multiple categories have accumulated strong price adjustment pressure at the same time. The pre-price adjustment trigger identifier uses high trigger strength categories as the core identifier to completely record the urgent distribution of price adjustments in the current batch.

[0020] Step S12: Perform inventory backlog rate detection on the product operation data to generate backlog rate distribution, obtain the competitor benchmark set through competitor market data, and establish a price adjustment monitoring rule table by adapting the backlog rate distribution to the competitor benchmark set according to the product category sales cycle.

[0021] In some embodiments, the step of performing inventory backlog rate detection and generating a backlog rate distribution based on the product operation data includes: identifying new product launch node records for each category based on the product operation data to construct a digestion rate curve; performing digestion rate drop segment analysis on the digestion rate curve to obtain a slow-moving slope value; performing a new-old product substitution association mapping based on the slow-moving slope value to obtain a slow-moving weight; and performing backlog intensity mapping based on the slow-moving weight and the slow-moving slope value to generate a backlog rate distribution.

[0022] Based on product operation data, new product launch node records for each category are identified to construct a digestion rate curve. The inventory inflow records for each category in the product operation data are the direct data source for identifying launch nodes. The core of determining a launch node lies in the relative increase in daily inventory inflow—a date record is triggered when the daily inventory inflow exceeds twice the average daily inventory inflow for that category. Records meeting this condition are considered launch nodes. Multiple launch nodes can occur for the same category within a statistical period, and each launch node independently triggers the digestion rate extraction process. After confirming a launch node, starting from the launch date of that category in the product operation data, the ratio of the outflow volume to the initial inventory volume for that category is extracted in each 7-day statistical window. This ratio is defined as the digestion rate for that window, reflecting the inventory clearance efficiency of that category during the corresponding period. The digestion rates for each window are arranged chronologically to form the digestion rate curve corresponding to that launch node. The initial value of the digestion rate curve is usually high, corresponding to the concentrated release of user demand during the new product launch phase. It gradually declines to a stable range over time and then tends to stabilize. When the digestion rate curve shows a clear inflection point after a stable period and continues to decline, the inflection point location and the slope of the decline constitute the core identification objects for analyzing the sudden drop in digestion rate. The numerical levels of several windows before the inflection point serve as a benchmark for the normal digestion rate of that product category, while the deviation of the values ​​in each window after the inflection point determines the identification boundary of the subsequent sudden drop segment. The set of digestion rate curves corresponding to all product categories and all shelf-ready nodes constitutes the complete input for extracting the sluggish sales slope value. The inflection point location and the decline range of the digestion rate curves jointly determine the identification boundary of the subsequent sudden drop segment analysis.

[0023] For example, the step of analyzing the digestion rate curve to obtain the slow-moving slope value includes: identifying replenishment nodes and locating the section where sales have not recovered in the digestion rate curve; identifying the duration of slow-moving inventory in the section where sales have not recovered to generate a slow-moving feature set; analyzing the time-varying changes in the slow-moving feature set to generate a slow-moving intensity parameter; and performing benchmark calibration based on the slow-moving intensity parameter and the section where sales have not recovered to generate a slow-moving slope value.

[0024] The inventory turnover rate curve is used to identify replenishment points and pinpoint areas where sales have not recovered. The increase in daily inventory relative to the average of the past 30 days is the core indicator for identifying replenishment points. Dates exceeding 1.8 times the threshold are automatically marked as replenishment points. Replenishment is usually accompanied by a short-term rebound in sales. A period in which the inventory turnover rate does not recover within two consecutive statistical windows after a replenishment point is considered a period where sales have not recovered. During this period, inventory continues to accumulate while sales are not effectively stimulated. If the inventory turnover rate continues to decline after replenishment and no recovery is seen in two statistical windows after the replenishment point, it indicates that the sluggish sales of this product category are not due to short-term shortages, but rather a structural decline in market demand after a specific point. Replenishment not only fails to improve the inventory situation but further exacerbates the backlog pressure. Conversely, if the inventory turnover rate increases significantly in the second statistical window after replenishment, it does not trigger the determination of a period where sales have not recovered. The difference in the shape of the two types of replenishment results in the inventory turnover rate curve is the direct basis for determining the boundary of a period where sales have not recovered. For product categories without replenishment points in the digestion rate curve, the equivalent starting point is the window where the digestion rate is below 50% of the category's historical average for three consecutive statistical windows. This equivalent starting point extends to the window before the digestion rate recovery window, serving as the termination boundary for the sales non-recovery segment. This equivalent starting point mechanism ensures that naturally slow-moving product categories without replenishment can still be included in subsequent analysis without being missed due to missing replenishment records. Multiple sales non-recovery segments may appear for the same product category in the digestion rate curve. Each segment independently records its start and end window positions. The start and end point information of multiple sales non-recovery segments are collectively incorporated into the subsequent process for identifying the duration of slow-moving inventory.

[0025] For segments where sales have not recovered, a feature set of slow-moving inventory is generated by identifying the duration of slow-moving inventory. The time span of each segment in the slow-moving inventory segment is measured by the number of statistical windows. The number of windows is multiplied by 7 days to obtain the number of days in the segment. 21 days (i.e., three complete windows) is the minimum entry threshold for a valid slow-moving inventory segment. Short-term segments in the slow-moving inventory segment that are shorter than this time span are considered as periodic fluctuations and are excluded. This design ensures that the slow-moving inventory determination has sufficient time span to support it. For valid slow-moving inventory segments in the slow-moving inventory segment that pass the entry threshold, the slow-moving inventory description of the segment is composed of three values: duration, average digestion rate within the segment, and lowest digestion rate within the segment. The longer the duration, the longer the inventory clearance ability has been hindered. The lower the lowest digestion rate within the segment, the worse the clearance efficiency during the most severe period of slow-moving inventory. The set of descriptive entries for all effective slow-moving segments constitutes the slow-moving characteristic set. When multiple effective slow-moving segments exist within the same product category in the slow-moving characteristic set, the entries for each segment are arranged in chronological order. The dense appearance of multiple segments indicates that the product category repeatedly falls into a slow-moving state throughout the statistical period, and inventory backlog has cyclical characteristics. For product categories that repeatedly enter a sales failure-to-recovery segment after the promotional period ends, if the duration of each segment gradually increases and the minimum digestion rate within the segment decreases accordingly, the intensity of slow-moving described by each entry in the slow-moving characteristic set shows an increasing trend. This reflects a continuous contraction in market demand rather than a short-term fluctuation, which is significantly different from the normal decline pattern of natural inventory digestion after holidays in the chronological structure of the slow-moving characteristic set. When the duration of the slow-moving ...

[0026] The sluggishness characteristic set is analyzed for time-varying changes to generate a sluggishness intensity parameter. Trend judgment is achieved through longitudinal comparison of adjacent sluggishness segments on the time axis of the sluggishness characteristic set—focusing on tracking the directional changes in duration and minimum digestion rate between segments. When the duration of sluggishness in the sluggishness characteristic set increases segment by segment and the minimum digestion rate decreases segment by segment, it is judged as a deteriorating trend; conversely, it is judged as an improving trend. When there is no significant monotonic change, it is judged as stable. The sluggishness intensity parameter of the category with a deteriorating trend is determined by the weighted sum of the average increase in duration and the average decrease in minimum digestion rate. Each of the two changes is taken as the normalized value of the entire category and weighted at 0.5. The weighted sum reflects the comprehensive degree of the deterioration rate of sluggishness in that category. For product categories where the duration and minimum digestion rate show consistent direction and significant variation across adjacent segments, the deterioration trend is readily apparent in the comparison of slow-moving characteristic set items, and the corresponding slow-moving intensity parameter is at a high level among all product categories. For product categories where neither the duration nor the minimum digestion rate shows significant directional change between adjacent segments, the slow-moving intensity parameter is the normalized value of the average minimum digestion rate within the slow-moving characteristic set segment. The difference in slow-moving intensity parameters between the two types of product categories fully reflects the differentiating role of trend directionality in assessing backlog risk. Product categories with improving trends are assigned lower slow-moving intensity parameters, while product categories with stable trends use the normalized value of the average minimum digestion rate within the slow-moving characteristic set segment as the slow-moving intensity parameter. For product categories with only a single effective slow-moving segment in the slow-moving characteristic set, a trend judgment cannot be formed; therefore, the normalized value of the minimum digestion rate of that segment is directly used as the slow-moving intensity parameter, resulting in a relatively conservative parameter estimate for single-segment product categories during the intensity calibration phase.

[0027] Based on the sluggishness intensity parameter and the sales recovery period, a benchmark calibration is performed to generate the sluggishness slope value. The original rate of change is calculated by dividing the absolute value of the difference between the digestion rate at the beginning and end of the sales recovery period by the number of days the period lasts, reflecting the average rate of decline in inventory digestion capacity during that stage. The sluggishness slope value K_s is then adjusted for trend by introducing the sluggishness intensity parameter P_s, determined by the formula K_s=R×(1+P_s), where R is the original rate of change in the sales recovery period, P_s is the normalized value of the sluggishness intensity parameter, and K_s incorporates the trend deterioration information into the original rate, making the calibration slope of categories with obvious deterioration trends significantly higher than the original rate of change. For categories with stable or improving trends, K_s is smaller in difference from the original rate due to the lower P_s. The maximum value of K_s across all effective periods for each category is taken as the representative sluggishness slope value for that category. This maximum value selection method ensures that the representative sluggishness slope value covers the most severe single sluggishness event in the category's history. For product categories with multiple sales stagnation periods during the statistical period, as the deterioration trend intensifies, the P_s value in the later periods continues to rise, and K_s increases accordingly. The representative stagnation slope value is the maximum K_s in the later periods, which fully captures the intensity of the most severe stagnation event for this product category. For product categories with similar original rate of change R values ​​but no obvious monotonic direction between the duration of stagnation feature sets and the lowest digestion rate, the stagnation intensity parameter P_s is lower due to the trend being judged as stable, and K_s is significantly smaller than the former. In the backlog intensity mapping stage, the difference in backlog intensity between the two types of product categories reflects the effective introduction of trend information by the benchmark calibration, so that the stagnation slope value truly reflects the actual stagnation intensity of each product category rather than just relying on a rate snapshot of a single period.

[0028] The slow-moving slope value is used to map the substitution relationship between new and old models in a product category to obtain the slow-moving weight. Before being mapped to the backlog intensity, the slow-moving slope value needs to be corrected for substitution association—new product launches and slow-moving old models often have a direct substitution relationship. The decreased sales rate of old models after the launch of new models is considered active substitution rather than actual backlog. If the slow-moving slope value of old models is directly mapped as backlog risk without distinction, normal product iteration will be misjudged as inventory backlog requiring price reduction intervention. In product operation data, if a new model is launched within 14 days before and after a sharp drop in the sales of old models in the same product category, it is considered a new-old model substitution event. During the period covered by the substitution event, the slow-moving slope value of old models is assigned a lower slow-moving weight, reflecting that the slow-moving portion during that period is due to active substitution rather than a contraction in market demand. For product categories where new models were launched within 14 days before and after a sharp price drop, and market purchasing behavior generally shifted towards the new models, this is a typical scenario of new-old model substitution. The slow-moving slope value of the old model corresponds to a lower slow-moving weight, significantly reducing the contribution of the old model's slow-moving inventory to the inventory risk assessment. For product categories where no new models were launched before or after the sharp price drop, and the decline in the sales rate was entirely due to shrinking demand, the slow-moving slope value corresponds to the highest slow-moving weight. The significant difference in slow-moving weight between the two categories accurately distinguishes the essential difference between substitution-induced slow-moving inventory and demand-shrinkage-induced slow-moving inventory. For product categories corresponding to sharp price drops without new-old model substitution, all values ​​of the slow-moving slope value participate in the subsequent inventory intensity mapping, maintaining a high slow-moving weight. The slow-moving weight value is set from 0.3 to 1.0, with a lower slow-moving weight corresponding to a higher degree of substitution event coverage. For sharp price drops without substitution events, the slow-moving weight is 1.0.

[0029] The backlog intensity mapping is performed based on the unsold inventory weight and unsold inventory slope value to generate the backlog rate distribution. The three-dimensional product model I=S×W_s×(1+D_norm) integrates the unsold inventory slope value (S), the proactive substitution correction (W_s), and the inventory age deviation (D_norm) into a single backlog intensity value. S is taken as the representative unsold inventory slope value of the category, W_s is taken as the unsold inventory weight of the corresponding category, and D_norm ranges from 0 to 1, determined by the formula D_norm=min(1,max(0,(actual inventory age-category standard turnover days) / category standard turnover days)). The greater the inventory age deviates from the standard turnover days, the closer D_norm is to 1, and the larger the backlog intensity product coefficient is. The product of the three factors comprehensively reflects the joint contribution of the unsold inventory rate, proactive substitution correction, and inventory age deviation to the backlog risk. Category categories with a large S, high W_s, and D_norm close to 1 typically correspond to situations where there has been sluggish sales for several consecutive statistical periods, frequent product iterations by competitors, and inventory age several times the standard turnover days. The combined effect of these three factors results in a high inventory intensity I in the inventory rate distribution. Category categories with similar sluggish sales slopes but lower sluggish sales weight W_s due to simultaneous new product launches, and a relatively reasonable inventory age leading to a lower D_norm, ultimately have significantly lower inventory intensity than the former. The numerical difference between these two scenarios in the inventory rate distribution reflects the combined moderating effect of substitution correlation correction and inventory age factors on inventory intensity assessment. After normalization across all categories, the inventory intensity I is mapped to the 0-1 range. The normalized inventory intensity of each category is arranged in order of category number to form the inventory rate distribution. High values ​​in the inventory rate distribution correspond to categories with large sluggish sales slopes, non-substitutional causes, and significantly higher inventory ages, while low values ​​correspond to normal digestion rates or temporary sluggish sales due to proactive substitution. The backlog rate distribution and the backlog intensity values ​​for each product category form the quantitative basis for matching the product category's sales cycle with competitor benchmarks.

[0030] Competitive benchmark sets are obtained through competitor market data. Historical selling prices of competitors in each category within the competitor market data are the direct source for extracting competitor price benchmarks. The extraction logic prioritizes the representativeness of regular prices—after removing promotional markers from the competitor market data, the median regular selling price of competitors in each category is used as the benchmark value. The median has a natural resistance to extreme low-price dumping behavior and will not lower the overall benchmark due to aggressive price reductions by individual competitors. Competitive price benchmarks for each category are combined with the corresponding category's backlog rate distribution to form differentiated benchmark tolerance ranges. Categories with high backlog intensity in the backlog rate distribution correspond to narrower competitor benchmark tolerance ranges. The narrowing of the tolerance range is positively correlated with the backlog intensity value, ensuring that the categories with the most severe backlog receive the most sensitive price tracking protection. Categories with fewer than three competitors in the competitor market data are marked as sparse competitor categories. The price benchmark for sparse competitor categories is replaced by the platform's historical average price to avoid excessive distortion of the benchmark by abnormal pricing of a single competitor. In cases where the number of competing products is extremely small, and some of them employ a persistently aggressive low-price strategy, directly using the average price of a few competing products as a benchmark would severely underestimate the normal selling price level of that category. The platform's historical average price substitution mechanism can bring the benchmark value back to the true market price center, and the tolerance range is also expanded accordingly to offset the uncertainty of estimation from a single data source. Sparse competitor labeling triggers additional tolerance range expansion processing simultaneously during differentiated benchmark matching. The competitor price benchmarks for all categories and the corresponding tolerance range sets constitute the competitor benchmark set. Each category entry in the competitor benchmark set includes three items: the price benchmark value, the upper and lower bounds of the tolerance range, and the number of competitors. The category coverage of the competitor benchmark set is consistent with the number of categories involved in the competitor market data, ensuring that subsequent differentiated benchmark matching has corresponding entries available for each category.

[0031] In some embodiments, the step of adapting the backlog rate distribution to the competitor benchmark set according to the product category sales cycle to establish a price adjustment monitoring rule table includes: dividing the backlog rate distribution into time periods according to the product category sales cycle to generate cycle rate groups; performing differential benchmark matching between the cycle rate groups and the competitor benchmark set to generate a tolerance critical benchmark; setting segmented price adjustment boundaries based on the tolerance critical benchmark; and performing interval mapping and arrangement of the segmented price adjustment boundaries to generate a price adjustment monitoring rule table.

[0032] The backlog rate distribution is divided into periodic rate groups based on the sales cycle of each product category. The backlog intensity value of each product category in the backlog rate distribution needs to be combined with the sales rhythm of the category itself to form an effective basis for segmentation. The sales cycle is estimated by anchoring the average interval between two adjacent sales peaks in historical transaction records. The shorter the number of days, the faster the turnover rhythm of the product category. Based on this, all product categories are divided into three groups: short cycle (less than 30 days), medium cycle (30 to 90 days), and long cycle (more than 90 days). The backlog intensity value of each product category in the backlog rate distribution is grouped according to the corresponding cycle to form the rate subset of each group. The three periodic rate subsets are labeled by group to form the overall periodic rate group. The backlog intensity values ​​for short-cycle groups are generally low. Connected vehicle fast-moving consumer goods (FMCG) accessories have a fast inventory turnover rate, and even if there are periods of slow sales, categories with fast sales pace can relatively easily digest inventory through short-term price adjustments. The duration and intensity of a single backlog event are usually lower than those for slow-moving categories. The backlog intensity values ​​for long-cycle groups are generally high. Slow-moving categories such as body modification parts and custom seat parts often experience prolonged periods of slow sales, with a single inventory clearance cycle spanning multiple statistical windows. Short-term intervention has limited effectiveness and requires more precise price adjustments. The backlog intensity values ​​for medium-cycle groups are usually between the two. Automotive electronics categories are mostly concentrated in this group. Affected by both product iteration cycles and seasonal demand fluctuations, the distribution pattern of backlog intensity shows significant intra-category differences. The distribution characteristics of the three subsets within the cycle rate group serve as independent references for determining subsequent tolerance thresholds. Cross-group comparisons within the cycle rate group are not used as a reference for benchmark matching. The category coverage of the cycle rate group is consistent with the number of categories involved in the backlog rate distribution.

[0033] Differentiated benchmark matching is performed on the cycle rate group and the competitor benchmark set to generate the tolerance threshold benchmark. Determining the comprehensive tolerance threshold value requires consideration of two dimensions: the backlog intensity distribution characteristics of the cycle rate group subset provide a reference for inventory pressure, and the upper and lower bounds of the tolerance range for the corresponding product category in the competitor benchmark set provide a reference for price competition. Cross-matching of these two dimensions forms the differentiated threshold value for each product category. For the short-cycle group, the tolerance threshold benchmark primarily references the upper bound of the competitor benchmark set's tolerance range. For every 0.1 increase in backlog intensity, the tolerance threshold benchmark narrows downward by 8% of the corresponding tolerance range width, reflecting the need for a faster response to competitor price changes when backlog intensity increases. For the long-cycle group, the tolerance threshold benchmark narrowing coefficient is lowered to 5%. Because the price response rhythm of slow-moving products is relatively relaxed, excessively narrowing the tolerance range would lead to overly frequent price adjustments, affecting price stability. For product categories with continuously increasing backlog intensity in the backlog rate distribution, the tolerance threshold benchmark continues to narrow after differential matching between the cycle rate group and the competitor benchmark set. The allowable deviation space between the current price and the competitor benchmark is correspondingly compressed, and any further price reduction by competitors may trigger a price adjustment signal. For product categories with similar backlog intensity but belonging to a long-cycle subset, the narrowing of the tolerance threshold benchmark retains some room for observation due to the smaller coefficient, and it is not advisable to initiate price adjustments too early. The difference in the tolerance threshold benchmark values ​​between the two categories reflects the structural impact of the sales cycle on the sensitivity of price adjustment responses. For product categories with sparse competitors in the competitor benchmark set, the tolerance threshold benchmark is extended by an additional 10% during matching to offset the risk of misjudgment caused by the uncertainty of benchmark estimation. For product categories with sufficient competitor information, the tolerance threshold benchmark is not extended further. The tolerance threshold benchmark values ​​for all categories are arranged by category number to form the tolerance threshold benchmark. Categories with lower values ​​indicate that the current backlog pressure is high and competitor prices are close to the tolerance boundary.

[0034] The price adjustment boundaries are set based on the tolerance threshold. Centered on the overall category average of the tolerance threshold, prices below one standard deviation of this average are classified as urgent price adjustments; prices between one and two standard deviations of this average are classified as regular monitoring; and prices above one standard deviation of this average are classified as lenient observation. Each of these three categories corresponds to different boundary widths and trigger sensitivity configurations. The urgent price adjustment segment has the narrowest boundary range, with the upper bound being the tolerance threshold value and the lower bound being 10% below it. This minimizes the price deviation required to trigger a price adjustment, and corresponds to the category with the highest price monitoring frequency. The regular monitoring segment's boundary range extends 15% above and below the tolerance threshold value, with moderate trigger conditions; price adjustment suggestions are pushed out in a pending-approval manner. The lenient observation segment has the widest boundary range, extending 25% above and below; a significant price deviation is required to trigger a price adjustment signal. This segment primarily records price deviation events and does not proactively push price adjustment suggestions. Categories with a tolerance threshold falling below one standard deviation of the overall category mean typically exhibit both long-term sluggish sales and recent continuous downward adjustments in competitor benchmarks. These categories have extremely narrow price adjustment boundary ranges, triggering adjustments even with a slight dip below the lower limit, resulting in the highest level of price monitoring density. Categories with a tolerance threshold falling into a more lenient observation range typically have ample inventory and stable competitor benchmarks in the near term. Their boundary ranges extend significantly to both sides, requiring substantial price deviations to trigger adjustments, effectively preventing frequent price adjustments from interfering with price stability. Categories in the regular monitoring range cover the largest number of categories, forming the core of the price adjustment monitoring system. The gradient differences in boundary width among these three categories fully reflect the differentiated needs for price adjustment response sensitivity due to inventory pressure and competitor background. These three rules are executed sequentially at the category level, with the resulting boundary values ​​for each category varying according to its respective segment. The sum of all category boundary values ​​forms the segmented price adjustment boundaries.

[0035] The segmented price adjustment boundaries are mapped and arranged to generate a price adjustment monitoring rule table. The three boundary values ​​for each category within the segmented price adjustment boundaries are transformed into executable rules after rule semantic conversion. Each executable rule includes five elements: category number, segment label, price upper and lower bounds, trigger condition judgment logic, and corresponding response action. After the three boundary values ​​are transformed into rule semantics one by one, the monitoring trigger logic for each category can directly drive the system to execute. The response action is configured differently according to the segment label: an urgent price adjustment segment trigger immediately pushes a price adjustment suggestion; a regular monitoring segment trigger generates a pending price adjustment reminder; and a lenient observation segment trigger only records the price deviation event without pushing proactive suggestions. Within the segmented price adjustment boundaries, the same category may experience segmentation changes due to updates in the backlog rate distribution over different statistical periods. During the interval mapping arrangement, a time-series annotation is added to the historical segmentation trajectory of each category. Categories whose segmentation has migrated from the lenient observation segment to the urgent price adjustment segment are marked as upgraded categories in the price adjustment monitoring rule table, indicating that the backlog risk of this category has significantly worsened recently. When the proportion of categories requiring urgent price adjustments significantly increases compared to the previous period, and upgraded categories are concentrated in the same sub-category group, the response priority of the price adjustment monitoring rule table for the corresponding category group increases accordingly. The phased changes in the proportion of categories requiring urgent price adjustments are a systemic signal measuring the current overall urgency of inventory backlog; a continuously rising proportion indicates that overall backlog pressure is entering a concentrated release phase. The monitoring trigger rules for all categories are arranged in order of category number to form the price adjustment monitoring rule table. The proportion of categories requiring urgent price adjustments in the price adjustment monitoring rule table reflects the current overall urgency of inventory backlog.

[0036] Step S13: Based on the pre-adjustment trigger identifier, the weight allocation of the price adjustment monitoring rule table is adjusted to generate an operation control rule set. The operation control rule set is used to conduct a check on the competitor's price-following response speed exceeding the limit to determine the intervention threshold node.

[0037] In some embodiments, the step of generating an operational control rule set by weighting the price adjustment monitoring rule table based on the pre-price adjustment trigger identifier includes: performing a distribution analysis of the collection dwell time of the pre-price adjustment trigger identifier to locate the price adjustment gap category; performing weight configuration matching based on the price adjustment gap category to generate dwell weight parameters; performing priority gradient sorting on the dwell weight parameters to determine the gap priority parameters; and performing rule priority rearrangement based on the gap priority parameters and the price adjustment monitoring rule table to generate an operational control rule set.

[0038] For example, the step of analyzing the collection dwell time distribution of the pre-adjustment trigger identifier to locate the price adjustment gap category includes: analyzing the collection dwell time distribution of the pre-adjustment trigger identifier to obtain dwell distribution features; identifying dwell time peaks based on the dwell distribution features to generate dwell peak groups; mapping the distribution characteristics of the dwell peak groups to construct a dwell gap attribute set; and performing gap association matching based on the dwell gap attribute set to locate the price adjustment gap category.

[0039] The distribution characteristics of the dwell time in favorites are obtained by analyzing the pre-price adjustment trigger identifier. The favorite behavior data corresponding to each trigger category of the pre-price adjustment trigger identifier is included in this step as the analysis object. Categories other than those triggered by the pre-price adjustment trigger identifier are not included in this dwell time distribution analysis. Each unsold record in this part of the favorite behavior data is marked with a dwell time field. The field value is the difference between the data extraction date and the date of the favorite behavior. After summarizing, the number of records in each of the four intervals of 0 to 7 days, 8 to 30 days, 31 to 90 days and more than 90 days are counted. The four interval percentage sequence corresponding to each trigger category of the pre-price adjustment trigger identifier constitutes the dwell time distribution outline of the favorite behavior of that category. For categories with a high proportion of short-cycle intervals, users make quick decisions after adding items to their favorites, with minimal price resistance. Users typically complete their purchase decisions shortly after adding items to their favorites, resulting in a clear short-end concentration in the four-interval distribution. Conversely, for categories with a prominent proportion of long-cycle intervals, users remain in a wait-and-see state for extended periods, exhibiting significant price resistance. When the proportion of intervals exceeding 90 days is significantly higher than the average level of categories triggering pre-price adjustments in the same batch, the inability of users in this category to make purchase decisions over a long period is clearly evident in the retention distribution characteristics, showing a significant long-end concentration in the four-interval distribution. Both types of distribution profiles trigger high-value signals in different intervals during the peak identification phase. Short-end dominant categories and long-end dominant categories exhibit drastically different user decision-making cycle characteristics in the favorites data. The difference in the shapes of these two profiles directly determines the peak interval landing point and the distribution pattern of values ​​exceeding multiples in subsequent peak determination. The set of retention distribution features for all triggered categories of pre-adjustment triggers fully covers the temporal structure of user price resistance behavior under the background of the rhythm shift of the current batch of competitors. The vector of retention distribution features for each category records the distribution density of the corresponding category's collection retention in different time periods.

[0040] Peak duration identification is implemented based on retention distribution characteristics to generate retention peak groups. The criterion for determining peak intervals is: the proportion of a certain interval in the retention distribution characteristics of a category exceeds twice the standard deviation of the mean of the entire batch in that interval. This threshold design ensures that only abnormally concentrated distribution patterns in the retention distribution characteristics will trigger peak labeling, eliminating interference from normal distribution differences between categories. Each category retains a representative peak interval. When multiple intervals exceed the threshold simultaneously, the interval with the largest deviation from the mean is selected. The representative peak interval identifier and the corresponding deviation multiple together constitute the peak entry for that category. The set of peak entries for all categories with peak intervals constitutes the retention peak group. Categories where the proportion of each interval in the retention distribution characteristics does not exceed twice the standard deviation of the mean do not generate peak entries, indicating that the retention distribution of this category is uniform and without obvious abnormal concentration, and therefore is not included in the peak group. Categories with peak retention periods concentrated between 31 and 90 days indicate that users abandoned their purchases after a period of observation. Since prices are close to the purchase threshold, moderate price adjustments may help revive these lost conversions. Categories with peak retention periods exceeding 90 days indicate more persistent price resistance; users have kept products in their favorites for extended periods without making a purchase. Relying solely on price adjustments will have limited conversion effects, requiring combined promotional incentives. Categories with peak retention periods of 31-90 days and those exceeding 90 days are separately included in the retention peak group and labeled with different peak ranges. For mid-term peak categories, users' observational psychology is closer to conversion decisions, while for long-term peak categories, a deeper price trust barrier exists. The two categories are clearly distinguished in their peak range labeling, and the different peak ranges directly affect the subsequent determination of the attribute type of the retention gap attribute set.

[0041] A distribution characteristic mapping was performed on the peak retention groups to construct a retention gap attribute set. Information from two dimensions was integrated: the time period of the peak determines whether the stagnation is medium-term or long-term, thus defining the gap attribute type; the magnitude of the exceedance measures the degree of concentration. The joint judgment result of these two dimensions was written into the retention gap attribute set. Categories in the peak retention group with peak periods of 31 to 90 days were labeled as medium-term price stagnation, and categories with peak periods exceeding 90 days were labeled as long-term price stagnation. Categories in the peak retention group whose exceedance values ​​exceed three standard deviations above the average of all categories were marked with a severe stagnation label in the retention gap attribute set, indicating that the price stagnation level of this category is the most prominent among the categories triggering pre-price adjustments in the current batch. Severe stagnation markers typically appear in subcategories where competitors maintain high prices for extended periods while the product itself fails to adjust accordingly. This occurs when the product's price remains unchanged during the period of sustained price stagnation in competitors, resulting in a large accumulation of user records of items repeatedly added to favorites but never purchased. These categories simultaneously carry both long-term stagnation attributes and severe stagnation markers in their retention gap attribute set. This dual overlay of long-term stagnation attributes and severe stagnation markers comprehensively records the combined accumulation of competitor price stagnation and user purchase stagnation in this category. The coexistence of these dual markers is one of the main conditions for prioritizing core members in the subsequent gap association matching stage. All retention peak group categories are mapped using distribution characteristics to form a retention gap attribute set. In addition to carrying the gap attribute type and severe stagnation marker status identified in this step, each category entry in the retention gap attribute set also synchronously attaches the cumulative trigger strength value of that category as a rhythm offset association field, based on the category number and the corresponding entry of the pre-adjustment trigger identifier. This allows the retention gap attribute set entries to simultaneously contain both retention-side attributes and rhythm-side strength information.

[0042] Based on the gap attribute set, gap correlation matching is performed to locate price adjustment gap categories. Correlation matching is based on the joint judgment of each field within the gap attribute set entry—categories that simultaneously meet the attribute field and trigger strength field criteria within the gap attribute set entry undergo dual verification. The cross-verification of the two types of fields within the gap attribute set entry constitutes the core support for price adjustment intervention. Categories with severe obstruction labels and high correlation trigger strength field values ​​in the gap attribute set are prioritized as core members of the price adjustment gap category. Users of these categories have long added them to their favorites but have not yet purchased them, and the price rhythm of competing products provides a proactive price adjustment window at this time. The superposition of the two types of fields within the gap attribute set entry provides the price adjustment intervention basis with support from both user-side demand and market-side timing. Categories without severe obstruction labels in the gap attribute set but with moderate correlation trigger strength fields are included as ordinary members of the price adjustment gap category through ordinary matching. Compared with core members, the strength of both criteria is moderate, placing them in the second-level gap gradient in the gap priority parameter ranking. Categories with missing or zero trigger strength fields for the retention gap attribute cluster will not generate price adjustment gap category records. Categories that are only based on collection retention without rhythmic offset support do not currently constitute price adjustment gaps. Single-dimensional signals are insufficient to trigger the confirmation process for price adjustment gap categories. This design avoids categories with abnormally concentrated collection behavior but no offset in the rhythm of competing products being mistakenly included in the price adjustment gap system.

[0043] Based on the price adjustment gap category, a weighted matching process is used to generate retention weight parameters. A weighted summation model uses two features to construct the initial weights: the proportion of long-term retention intervals (normalized weight 0.6) and the median retention duration (normalized weight 0.4). The former, with its higher weight, reflects the stronger indicative effect of prolonged non-purchase behavior on the degree of price stagnation. The initial retention weight parameters are determined by the formula W_r0 = 0.6×L_norm + 0.4×M_norm, where L_norm is the normalized value of the proportion of long-term retention intervals, M_norm is the normalized value of the median retention duration, and the weighted sum W_r0 serves as the initial retention weight parameter value for this category. The original retention weight parameter W_r0 is normalized to the range of 0.2 to 1.0 after being normalized for all price adjustment gap categories. The normalized value is denoted as W_r and serves as the final value of the retention weight parameter. The lower bound is set to 0.2 to ensure that each price adjustment gap category obtains basic response resources in the rule reordering. When each category entry of the retention weight parameter is generated, the association trigger strength field carried by the upstream price adjustment gap category is retained as the internal constraint field of the entry. The retention weight parameter entry is composed of the normalized weight main field and the association trigger strength constraint field. Among the price adjustment gap categories, the collection records are mainly long-term, with more than half of the records having extremely long retention periods. The phenomenon of users adding items to their collections for a long time but never purchasing them is the most prominent, corresponding to a very high proportion of long-term retention intervals. The original weight is at the high end of the interval, and the retention weight parameter after normalization ranks among the top in all price adjustment gap categories. Categories with collection records concentrated in a relatively short period and a short purchase blocking period have a lower proportion of long-term retention intervals, a smaller median retention time value, and a correspondingly lower retention weight parameter value. Categories with concentrated medium-term retention fall between the two. The three categories form a clear gradient in the retention weight parameter distribution, and the differentiated impact of long-term and short-term blocking on weight allocation is fully reflected in the parameter distribution.

[0044] Priority gradient sorting is performed on the retention weight parameters to determine the gap priority parameters. Solely relying on the main field of the retention weight parameters for sorting has significant drawbacks—categories with insufficient rhythm offset accumulation may falsely occupy high positions due to prominent retention distribution. Therefore, the associated trigger strength constraint field within the retention weight parameter entries is used as the constraint dimension. The comprehensive priority score is determined by the formula Q_i = W_r × T_acc, where W_r is the normalized weight main field of the retention weight parameter entries, and T_acc is the associated trigger strength constraint field within the retention weight parameter entries. The product of these two values ​​constitutes the comprehensive priority score Q_i, effectively suppressing categories with a strong bias in one dimension and a weak bias in another. Categories with high retention weight parameter main field levels across all price adjustment gap categories, and also high retention weight parameter constraint field levels, rank highly in the comprehensive score due to the combined strength of these two values. This dual-dimensional prominence ensures they receive the highest priority adjustment in subsequent rule re-ranking. Categories with only high retention weight parameter main field levels but low retention weight parameter constraint field levels experience a lower comprehensive score due to the reduced constraint field levels, effectively suppressing single-dimensional prominence. After being grouped into different gradients, these two categories receive corresponding priority adjustments during rule re-ranking. The comprehensive priority scores are sorted in descending order and then divided into three gradients based on quantiles: the top 30% are assigned to the high-priority gradient, 30% to 70% to the medium-priority gradient, and below 70% to the low-priority gradient. The gradient boundaries are dynamically determined based on the number of price adjustment gap categories and score distribution, ensuring that the proportion of categories in each gradient remains stable across different statistical periods. High-priority gradient categories are recorded as Level 1 gaps in the gap priority parameter, medium-priority gradient categories are recorded as Level 2 gaps, and low-priority gradient categories are recorded as Level 3 gaps. The gap level label of each category and the comprehensive priority score together constitute the gap priority parameter entry.

[0045] Based on the gap priority parameter and the price adjustment monitoring rule table, the rule priority is rearranged to generate an operational control rule set. The level label of each category in the gap priority parameter is the direct input of the mapping rule. The mapping rule is executed differently according to the gap priority parameter level: the rule priority of the first-level gap category is increased by two levels based on the original price adjustment monitoring rule table, the second-level gap category is increased by one level, and the priority of the third-level gap category remains unchanged. The three-level processing logic ensures that categories with both prominent evidence can overcome the response lag of the original segment configuration. In the price adjustment monitoring rule table, categories that were originally in the urgent price adjustment segment and had reached the highest priority will no longer be upgraded after reordering, to avoid excessive concentrated consumption of operational resources due to high-frequency triggering. Category 1 gap categories that were originally in the lenient observation segment can be upgraded to a response priority comparable to the regular monitoring segment after two tiers of upgrades. The static segmentation configuration is effectively corrected by dynamic adjustment. After the upgrade, the price monitoring frequency for category 1 gap categories is increased, and price adjustment suggestions are pushed out faster when the real-time price of this category triggers boundary conditions. The priority of category 3 gap categories remains unchanged, and dynamic adjustments only apply to category 1 and category 2 gap categories. The difference in priority handling between the two categories reflects the targeted correction effect of the gap priority parameter on the price adjustment monitoring rule table, ensuring that price adjustment resources are concentrated on categories with the most prominent double signal superposition. After all categories in the price adjustment monitoring rule table have completed priority reordering, the original segmentation labels and price boundary values ​​of each category are retained, and only the priority field is updated to form an operational control rule set. The category coverage of the operational control rule set is completely consistent with the price adjustment monitoring rule table.

[0046] In some embodiments, the step of determining intervention threshold nodes by checking the competitor's price-following response speed exceeding the limit for the operational control rule set includes: parsing the price-following response delay of each price adjustment action based on the operational control rule set to form a price-following delay sequence; identifying the delay threshold deviation of the price-following delay sequence to generate a delay deviation feature set; using the delay deviation feature set to filter competitor actions whose response speed exceeds the critical threshold to obtain an intervention configuration table; and extracting intervention threshold nodes by classifying the intervention intensity in the intervention configuration table.

[0047] Based on the analysis of the price adjustment action response delay of each price adjustment action using the operational control rule set, a price adjustment delay sequence is formed. The single price adjustment response delay is defined as the difference between the earliest time of a competitor's price change in the same product category and the time of the price adjustment action of this product. A positive difference indicates that the competitor followed suit after the price adjustment of this product, while a negative difference indicates that the competitor had already adjusted its price in advance. The distribution pattern of these two scenarios in the delay sequence directly reflects the degree of initiative of competitors in price adjustment for that product category. The single price adjustment response delay of all historical price adjustment actions for each product category in the operational control rule set is arranged in chronological order to form the price adjustment delay sequence for that product category. When no competitor follow-up records are detected for certain price adjustment actions in the operational control rule set within the statistical period, the corresponding delay record is set to the maximum upper limit of the statistical period days, ensuring that price adjustment actions without follow-up records do not lower the overall average of the price adjustment delay sequence. A persistently short price follow-up delay sequence indicates that competitors respond quickly during this period, resulting in a narrow window of opportunity for price adjustments for this product. For product categories with generally short price follow-up delay sequences during peak seasons, competitors are exceptionally sensitive to price changes, and the probability of a price adjustment initiated by this product during this period being quickly followed is significantly higher than in other periods. When the price follow-up delay sequence for the same product category lengthens significantly during the off-season, the monitoring frequency of competitors decreases significantly. The significant difference in delay distribution characteristics between peak and off-seasons indicates that the price adjustment advantage window for this product category is extremely narrow during peak season but relatively ample during off-season, making the off-season a better time to initiate proactive price adjustments. The distribution differences of the price follow-up delay sequence across different time periods provide a direct quantitative reference for the refined selection of price adjustment timing. Periods where historical price adjustment records for competitors are missing in the competitor market data correspond to the upper limit values ​​that need to be filled in the price follow-up delay sequence. Periods with dense upper limit value filling reflect that competitors generally make few price adjustments during these periods.

[0048] Delay threshold deviations are identified in the price-following delay sequence to generate a delay deviation feature set. The determination of ultra-fast response nodes relies on the statistical boundaries of the category's historical distribution, rather than a fixed duration threshold—a single price-following response delay falling below the historical mean minus two standard deviations is considered an indication that the competitor's response speed to this product's price adjustment action is far beyond normal levels, suggesting that the competitor may be conducting high-frequency, targeted price monitoring for this category. All ultra-fast response nodes in the price-following delay sequence for each category are extracted chronologically, recording the node's occurrence time, the corresponding price-following response delay value, and the standard deviation multiple from the historical mean. These three pieces of information constitute the deviation description entry for that node. The set of deviation description entries for all ultra-fast response nodes across all categories constitutes the delay deviation feature set. Time periods with a high density of ultra-fast response nodes for the same category in the price-following delay sequence form a high-density deviation entry distribution in the delay deviation feature set, indicating that the competitor's price-following activity in this category remains high during that period, severely compressing the execution window for this product's price adjustment action. For product categories with significant seasonal characteristics, competitors typically intensify their price monitoring efforts in the lead-up to the peak season. This results in a high density of deviation entries in the time delay feature set before the peak season, a stark contrast to the sparse distribution of deviation entries during off-peak periods. This indicates a significantly higher risk of price adjustments being quickly followed before the peak season compared to the off-season. Operations teams should prioritize pushing price adjustments during periods of low competitor monitoring activity. Statistical periods with a generally large deviation standard deviation multiple in the time delay feature set suggest that competitors have concentrated on increasing their price-following response speed across multiple product categories during that period, and the distribution structure of the price-following delay sequence has narrowed overall within that period.

[0049] Intervention configuration tables are obtained by filtering competitor actions with response speeds exceeding a critical threshold using a latency deviation feature set. Instead of a fixed duration threshold, an adaptive critical threshold is constructed for each product category – the threshold is the historical average of the price-following latency sequence for each category minus two standard deviations. This ensures that categories with different competitor activity benchmarks can accurately capture unconventional price-following behavior, rather than being misjudged or missed by a uniform threshold. All competitor price adjustment records corresponding to entries in the latency deviation feature set with a deviation exceeding two standard deviations are extracted as over-limit competitor actions. The cumulative number of over-limit actions and the average over-limit latency for the same product category within the statistical period in the latency deviation feature set jointly reflect the price-following intensity of that competitor in that category. The summary records of over-limit competitor actions for each category, combined with the response priority level of that category in the operational control rule set, form a category intervention demand description. In the operational control rule set, intervention demands for first-level gap categories are marked as high intensity, second-level gap categories as medium intensity, and third-level gap categories as low intensity. Competitor combinations with a significantly higher number of exceedances within the statistical period and belonging to a category with a Level 1 gap are recorded with high-intensity intervention requirements in the intervention configuration table; combinations with fewer exceedances and belonging to a Level 3 gap are recorded with low-intensity intervention; categories with both high-intensity markings and numerous exceedance records fully record the concentrated manifestation of competitor price-chasing intensity in that category. The intervention requirement descriptions for all categories and the summary records of exceedance competitor actions are integrated to form the intervention configuration table. The rows of the intervention configuration table correspond to the category and competitor combination, and the columns correspond to the intervention intensity marking, the number of exceedances, and the average exceedance delay.

[0050] Intervention threshold nodes are generated by classifying intervention intensity in the intervention configuration table. Node levels are determined by a combination of two indicators: intervention intensity labeling and the percentage of instances exceeding the limit. Combinations with high intensity labels and exceeding the limit by more than 30% of all competitor actions in the category are upgraded to Level 1 nodes; combinations with medium or high intensity labels but less than 30% exceeding the limit are downgraded to Level 2 nodes; and low intensity labels correspond to Level 3 nodes. This three-level classification ensures that combinations with different levels of competitor price-following threat receive matching intervention response configurations. The triggering condition for each level of intervention threshold node is based on the average exceeding delay of the corresponding combination. When the real-time monitoring shows that the competitor's price-following response delay in the category is lower than 80% of the average exceeding delay, the current node is considered triggered. After triggering, a delay is applied to the timing of price adjustments for the corresponding category in the operational control rules set. The delay is largest for Level 1 nodes and smallest for Level 3 nodes. When multiple intervention threshold nodes of different levels appear in the intervention configuration table for the same category, each level of node is triggered independently and its own delay strategy is executed. The timing of price adjustments for categories with multiple levels of nodes is subject to the strictest constraints. Competitor combinations that consistently maintain an extremely fast price-following pace are marked with high intensity in the intervention configuration table and have a high percentage of exceeding the limit, generating a Level 1 intervention threshold node. When the price-following latency of this competitor shows a significant contraction on the same day, the Level 1 node is triggered. The operational control rule set applies the maximum delay to the price adjustment push for this category. The operations side then prioritizes completing the price adjustment during periods with lower competitor monitoring activity to prevent the price adjustment strategy from being quickly imitated and copied within a highly sensitive window. Competitor combinations that occasionally accelerate price-following and whose category has a Level 3 gap correspond to a Level 3 intervention threshold node. The delay range when triggered is limited, which avoids the risk of price-following in that particular instance without excessively affecting the overall efficiency of price adjustment timing. The gradient difference in the delay range configuration between Level 1 and Level 3 nodes ensures that competitor behaviors with different price-following threat intensities receive intervention levels commensurate with their respective strengths.

[0051] Step S14: Based on the intervention threshold node, perform threshold mapping on the operation control rule set to activate and output the price adjustment behavior sequence. Extract the car owner click-but-not-purchase feature from the price adjustment behavior sequence to generate the recommendation gap coefficient. Based on the recommendation gap coefficient, perform complementary product association recommendation trend monitoring on the price adjustment behavior sequence to generate sales compliance.

[0052] Specifically, the operation and control rule set is activated and outputs a price adjustment sequence based on intervention threshold nodes. The mapping process is based on node level—when a level 1 node is triggered, the price adjustment push for the corresponding category is subject to the maximum delay constraint; when a level 2 node is triggered, the delay is moderate; when a level 3 node is triggered, the delay is limited; categories that have not triggered any nodes are directly activated according to the original priority configuration of the operation and control rule set. The four processing paths jointly determine the actual execution time of the price adjustment action for each category. Threshold mapping activation is based on the price trigger boundary of each category rule entry in the operation and control rule set. When real-time price monitoring detects that the trigger boundary condition for a certain category is met, it determines whether to immediately push a price adjustment suggestion or apply a corresponding level of delay based on the intervention threshold node constraint. After the delay ends, the adjustment is automatically pushed to ensure that the price adjustment action is not permanently shelved due to node constraints. When competitor rhythm shifts and inventory backlog pressures overlap during the same period, multiple product category push records appear concentratedly in the price adjustment behavior sequence. The delayed release nodes for Category 1 intervention levels fall around the end of the competitor's active window, consistent with the prediction of the peak competitor monitoring density based on the intervention threshold node. Product category push records that did not trigger any intervention nodes are relatively evenly distributed on the timeline, contrasting sharply with the concentrated release pattern of Category 1 intervention in the price adjustment behavior sequence. The difference in the distribution of push times for the two categories directly demonstrates the actual shaping effect of the threshold mapping activation mechanism on the overall price adjustment rhythm. All product category price adjustment suggestion push records generated within the statistical period are arranged chronologically, containing four elements: category number, suggested price adjustment range, push time, and corresponding intervention level. This set constitutes the price adjustment behavior sequence, and the temporal distribution of push times within the price adjustment behavior sequence reflects the density characteristics of the overall price adjustment rhythm within the current period.

[0053] In some embodiments, the step of extracting the "click-but-not-purchase" feature of car owners from the price adjustment behavior sequence to generate a recommendation gap coefficient includes: obtaining car owner click records and transaction records for each product based on the price adjustment behavior sequence; identifying conversion gaps in the car owner click records and transaction records to generate a gap feature set; determining gap category number groups by analyzing gap category attributes through the gap feature set; and generating a recommendation gap coefficient by associating and arranging the gap category number groups with the gap feature set.

[0054] The system retrieves driver click and transaction records for each product from the price adjustment behavior sequence. A 72-hour window is set as the behavior matching window duration. Click records generated within this period after the push notification are driver click records, while order records completed during the same period are transaction records. This window duration ensures comprehensive coverage for both impulsive purchases and delayed decision-making conversion paths. Driver click records include the clicked product number, click time, and source page label. Transaction records include the sold product number, transaction amount, and order time. Both types of records use the push notification record number from the price adjustment behavior sequence as the association key, ensuring that each click and transaction record can be traced back to the specific price adjustment trigger event. If a product category in the price adjustment behavior sequence has neither driver clicks nor transactions within 72 hours of the push notification, the corresponding click and transaction fields are left blank. A high percentage of blank fields indicates that the price adjustment push notification for that category has a weaker actual reach to users, possibly due to insufficient push channel coverage or the price adjustment not meeting user expectations. For product categories where car owners consistently click on price adjustments repeatedly and transaction records are almost nonexistent, the price adjustments failed to activate user browsing interest. The continued absence of both clicks and transactions suggests that the price elasticity of this product category is inherently low, and relying solely on price adjustments within the price adjustment sequence is unlikely to drive transaction conversion.

[0055] A gap feature set is generated by identifying conversion gaps in car owner click records and transaction records. The difference between the number of car owner click records and the number of transaction records after a price adjustment push is the core source of quantifying the conversion gap. The conversion gap rate G_gap is defined as the proportion of clicks that do not convert into transactions within the same behavior matching window, determined by the formula G_gap=(N_click-N_convert) / N_click, where N_click is the total number of clicks in the car owner click records, and N_convert is the number of orders that have completed payment in the transaction records. G_gap is only calculated when N_click>0. Push records with N_click of zero do not trigger gap detection. The value of G_gap ranges from 0 to 1. A category is judged as having a high gap when G_gap exceeds 1.5 times the historical average of the category, a category is judged as having a medium gap when G_gap is between 1.0 and 1.5 times, and a category is judged as having a gap when G_gap is below the historical average. The gap description entries for high-gap and medium-gap product categories record two core attributes: the distribution of click source pages reflects the entry scenario when conversion obstacles occur; the distribution of dwell time when abandoning purchase is the value of the time from when the user enters the product details page to when they leave, and the percentage of each of the three levels is statistically analyzed as less than 30 seconds, 30 to 120 seconds, and more than 120 seconds, reflecting the browsing depth when the user abandons the purchase. The set of gap description entries for all gap categories constitutes the gap feature set. In the gap feature set, categories with click sources concentrated on search results pages indicate that users reached them through active search but failed to achieve conversion due to price or product description. In the context of a combination of dense search clicks and sparse transactions under the background that the price is still higher than the average price of major competitors after price adjustment, the gap feature set clearly identifies the conversion obstacle scenario of this category. The persistently high G_gap indicates that the price gap has not been bridged, which is the core cause of the current conversion gap. Categories with click sources concentrated on recommendation positions indicate that there is a gap in the accuracy of recommendation matching. The distribution pattern of the two types of sources is recorded as a vector of the percentage of source pages in the gap feature set. The degree of difference of each component of the vector directly affects the attribute type determination in the attribute analysis stage of the gap category.

[0056] The gap category ID group is determined by analyzing the gap category attributes using the gap feature set. The click source and abandoned purchase dwell time distribution recorded in the gap feature set are the two core dimensions for gap attribute determination. Categories whose click sources are concentrated on the search results page are labeled as search conversion gap attributes, those whose click sources are concentrated on the recommendation slot are labeled as recommendation conversion gap attributes, and those with similar proportions of the two types of sources are labeled as mixed gap attributes. These three attribute labels reflect the differences in behavioral scenarios that hinder conversion for each category. The abandoned purchase dwell time distribution within the gap feature set is measured by the dwell time from entering the product details page to leaving. Abandonment records with a dwell time of less than 30 seconds are classified as quick browsing abandonment, while those with a dwell time of more than 120 seconds are classified as deep browsing abandonment. Deep browsing abandonment is more price-sensitive, and the corresponding category has a higher probability of successfully overcoming user decision-making obstacles through precise price adjustments than categories with quick browsing abandonment. Categories with a deep browsing abandonment rate exceeding 50% of all abandonment records in the gap feature set are classified as high price-sensitive gap categories, and their corresponding category IDs are included in the priority set of gap category ID groups; the remaining gap category IDs are included in the general set. Deep browsing abandons categories with dense records. Users repeatedly check specifications and prices but still choose to leave, indicating that the main obstacle to purchasing decisions is that the price does not meet expectations. The corresponding category numbers are included in the priority set and receive a bonus during the recommendation gap coefficient generation stage. Quick browsing abandons categories where purchasing obstacles are more due to product descriptions than price factors. These are included in the ordinary set. The category numbers of the gap category number group in the priority set and the ordinary set together cover all gap categories in the gap feature set.

[0057] The recommendation gap coefficient is generated by associating gap category ID groups and gap feature sets. The gap feature set provides quantitative information on conversion behavior for each category, while the gap category ID group provides the basis for price sensitivity grouping. From the matching results, three quantitative values ​​are extracted: the conversion gap rate G_gap recorded in the gap feature set, the percentage of deep browsing abandonment F derived from the dwell time distribution of the gap feature set, and the gap attribute type assigned A based on the click source distribution of the gap feature set. The recommendation gap coefficient C_i = 0.5 × G_norm + 0.3 × F_norm + 0.2 × A_norm is then weighted and integrated into a single coefficient. The numbers are defined as follows: G_norm is the normalized value of G_gap, F_norm is the normalized value of F, and A_norm is linearly mapped to the 0-1 interval according to the formula A_norm=(A-0.6) / 0.2. A is determined according to the gap attribute type: 0.8 for search conversion gap, 0.6 for recommendation conversion gap, and 0.7 for mixed gap. The conversion gap rate has the highest weight, reflecting the dominant role of overall conversion loss on recommendation priority. The second highest weight reflects the key influence of price sensitivity on the conversion potential of complementary products. The recommendation gap coefficient C_i is mapped to the 0-1 interval after normalization of all gap categories. The categories in the priority set of gap category number groups are given an additional 0.15 on top of C_i and truncated to the upper limit of 1.0 to ensure that the categories in the priority set are generally higher than the categories in the ordinary set in the distribution of recommendation gap coefficients. Categories with consistently high G_gap, dense deep browsing abandonment records, and a gap attribute indicating a search conversion gap all have relatively high levels of all three quantitative values. Their recommendation gap coefficient is among the highest of all gap categories, and their complementary product recommendation list allocation is the longest. Categories with medium G_gap and a deep browsing abandonment rate close to the threshold have a moderate recommendation gap coefficient and a shorter recommendation list. The recommendation gap coefficient for each category is a core parameter for recommendation weighting and directly determines the proportion of recommendation resources allocated to each category.

[0058] Based on the recommendation gap coefficient, the sales compliance of complementary product recommendation trends is generated by monitoring the price adjustment behavior sequence. The allocation of complementary product recommendation resources is directly driven by the recommendation gap coefficient. For gap categories, the product details page displays related connected car products with common usage scenarios (such as car navigation brackets, chargers, car charging cables, etc.): the recommendation list length for categories with a recommendation gap coefficient higher than 0.7 is 5 items, the length for categories with a recommendation gap coefficient between 0.4 and 0.7 is 3 items, and the length for categories with a recommendation gap coefficient lower than 0.4 is 1 item. The category with the most severe conversion stagnation receives the strongest complementary product association support. After the recommendation list is published, the monitoring window for the corresponding category of the price adjustment behavior sequence is extended to 96 hours after publication. The sum of the number of complementary product transaction records and the number of transaction records for the gap category within the monitoring window is defined as the joint sales volume. The percentage increase in joint sales volume relative to the sales volume of the same category in the baseline window 96 hours before the price adjustment behavior sequence is used as the sales improvement rate. A sales improvement rate exceeding 15% is considered compliant, between 5% and 15% is considered needing optimization, and below 5% is considered non-compliant. The compliance assessment results for each category, along with the sales improvement rate, constitute the sales compliance entries. All sales compliance entries are arranged chronologically to form the sales compliance score. When non-compliant entries in the sales compliance score are concentrated in a specific sub-category (such as typical complementary product groups like dashcams and memory cards), it indicates a discrepancy between the complementary product association logic of that category and the actual user purchase combination pattern. The co-purchase matching of the corresponding category in the association database needs to be re-examined.

[0059] Step S15: Based on the sales compliance and price adjustment behavior sequence, perform intervention recurrence cycle analysis to generate intervention recurrence cycle; based on the intervention recurrence cycle and operation control rule set, implement search term and transaction category deviation rate calibration to generate category operation profile; compare category operation profile with product operation data to generate intelligent operation evaluation results.

[0060] Specifically, intervention recurrence cycles are generated based on the analysis of sales compliance and price adjustment behavior sequences. The extraction of the effective intervention event set uses the push record number as the association key, filtering push records that have passed compliance judgment in the price adjustment behavior sequence. The median of the time interval between two adjacent effective intervention events is taken as the representative value of the intervention recurrence cycle for that category. The median method is robust to occasional abnormally long intervals and is not influenced by single long observation periods. When the number of effective intervention events for categories with sparse compliance records in sales compliance is less than two, the statistical period days are used as the upper limit estimate of the intervention recurrence cycle, with a low-frequency intervention label added, indicating that the effective intervention trigger frequency for that category is insufficient and a regularity has not yet been formed. For product categories with short intervention recurrence cycles, the stable intervals between adjacent effective intervention events closely match the push rhythm in the price adjustment sequence, indicating persistent inventory backlog pressure and repeated proactive price adjustments triggered by competitors' rhythms, resulting in stable intervention effectiveness. For product categories with long intervention recurrence cycles, the distribution of compliance records in sales compliance is scattered, with significant differences in adjacent intervals. The distribution of push times in the price adjustment sequence shows a clear asymmetry with the distribution of compliance nodes in sales compliance, indicating that effective price adjustments are relatively dispersed under the current operational rhythm, and the concentration of intervention resources needs to be improved. The accurate low-frequency intervention labeling accurately identifies the operational characteristic of insufficient intervention stability in this type of product category. The representative value of the intervention recurrence cycle and the low-frequency intervention labeling status together constitute the intervention frequency attribute description for each product category, which is directly referenced as a core field during the product category operational profile generation stage.

[0061] Based on the intervention reproduction cycle and operational control rule set, the deviation rate between search terms and transaction categories is calibrated to generate category operation profiles. The deviation rate P_dev is calculated using the formula P_dev = 1 - N_match / N_click_search, where N_match is the number of orders completed within 72 hours after a search click that are for the same category, and N_click_search is the total number of clicks on the search results page. P_dev is only calculated when N_click_search > 0. A higher P_dev indicates a more severe mismatch between search traffic and conversion categories, and a lower actual efficiency of search traffic resources in contributing to transactions. The calibration analysis weights of P_dev for each category are jointly determined by the intervention recurrence cycle and the operational control rule set. Categories with shorter intervention recurrence cycles have a higher base weight, while categories with low-frequency intervention have a lower base weight. The weights are then adjusted progressively upwards based on the response priority level of the corresponding category in the operational control rule set. Categories with first-level gaps receive the largest bonus, second-level gaps receive a moderate bonus, and third-level gaps receive no bonus. The weights determined by this dual basis are applied to the statistical summary of P_dev. The calibration results primarily reflect the search conversion status of categories with high-frequency intervention and high gap priority. Categories with shorter intervention recurrence cycles and higher P_dev exhibit a dual characteristic in the category operation profile: a coexistence of high-frequency intervention needs and search traffic misalignment. Categories with low intervention frequency but also high P_dev show a higher urgency for keyword optimization than for price adjustment intervention. The category operation profile records the intervention frequency attribute and search term deviation rate calibration value of each category using the category number as an index. At the same time, the response priority level of the corresponding category in the operation control rules set is used as the rule priority field, with the first-level gap category being used as the highest rule priority and the third-level gap category being used as the lowest rule priority.

[0062] Intelligent operational evaluation results are generated by comparing category operation profiles with product operation data. The primary verification dimension focuses on the consistency between the frequency of interventions and the trend of backlog intensity changes. When the category operation profile shows high-frequency interventions but the backlog intensity in the product operation data does not improve, the intervention effectiveness is judged to be low, indicating a structural deviation between the price adjustment range or timing and the actual sales demand. When the search term deviation rate in the category operation profile is high and the proportion of organic traffic transactions in the product operation data remains low, search traffic acquisition is judged to be ineffective, indicating that the keyword strategy needs systematic adjustment. The differences between categories are measured by the weighted average of the standardized deviations of the values ​​of each field in the category operation profile and the corresponding indicators in the product operation data. The weight is determined based on the variance of each field in the entire category. The larger the variance of a field, the higher its contribution to distinguishing differences between categories. The weight is equal to the ratio of the variance of that field to the sum of the variances of all fields. Categories with a weighted average exceeding the upper quartile of the entire category distribution are judged to be in urgent need of optimization, those between the upper quartile and the median are judged to be under routine monitoring, and those below the median are judged to be performing as expected. Category operation profiles show that categories with high-frequency interventions but persistent backlogs and significantly high search term deviation rates exhibit substantial gaps in both intervention effectiveness and search traffic generation. The intelligent operation assessment results classify these categories as requiring urgent optimization. Categories with moderate intervention frequency, steadily decreasing backlog intensity, and low deviation rates are considered to have met performance standards. The phased changes in the proportion of categories requiring urgent optimization across all categories in the intelligent operation assessment results record the overall improvement trajectory after the gradual implementation of operational intervention strategies. The assessment results for each category, key gap dimensions, and corresponding adjustment directions together constitute the intelligent operation assessment result items. All items are arranged according to assessment level, with categories requiring urgent optimization placed first for priority processing.

[0063] To implement the AI-based intelligent operation method for connected vehicle goods corresponding to the above method embodiments, and to achieve the corresponding functions and technical effects. See also Figure 2 , Figure 2 This application provides a structural block diagram of an AI-based vehicle-to-everything (V2X) intelligent goods operation system, which includes: Data acquisition module 201 is used to collect product operation data and competitor market data, and to perform price rhythm coupling analysis on the product operation data and competitor market data to generate a pre-adjustment trigger identifier; Inventory detection module 202 is used to perform inventory backlog rate detection on the product operation data to generate backlog rate distribution, obtain a competitor benchmark set through the competitor market data, and establish a price adjustment monitoring rule table by adapting the backlog rate distribution to the competitor benchmark set according to the product category sales cycle. The rule allocation module 203 is used to generate an operation control rule set by weighting the price adjustment monitoring rule table based on the pre-price adjustment trigger identifier, and to conduct a competitor price-following response speed exceeding limit verification to determine the intervention threshold node for the operation control rule set; Recommendation monitoring module 204 is used to perform threshold mapping on the operation control rule set based on the intervention threshold node to activate and output the price adjustment behavior sequence, extract the car owner click-not-purchase feature for the price adjustment behavior sequence to generate a recommendation gap coefficient, and perform complementary product association recommendation trend monitoring on the price adjustment behavior sequence based on the recommendation gap coefficient to generate sales compliance. The result output module 205 is used to perform intervention recurrence cycle analysis based on the sales compliance and the price adjustment behavior sequence to generate an intervention recurrence cycle, perform search term and transaction category deviation rate calibration based on the intervention recurrence cycle and the operation control rule set to generate a category operation profile, and compare the category operation profile with the product operation data to generate an intelligent operation evaluation result.

[0064] The aforementioned AI-based intelligent operation system for connected vehicle goods can implement the AI-based intelligent operation method for connected vehicle goods described in the above method embodiments. The options in the above method embodiments are also applicable to this embodiment and will not be detailed here. The remaining content of this application's embodiments can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0065] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

Claims

1. An AI-based intelligent operation method for connected vehicle goods, characterized in that, include: Collect product operation data and competitor market data, and perform price rhythm coupling analysis on the product operation data and competitor market data to generate a pre-adjustment trigger identifier; For the product operation data, perform inventory backlog rate detection to generate backlog rate distribution, obtain the competitor benchmark set through the competitor market data, and establish a price adjustment monitoring rule table by adapting the backlog rate distribution to the competitor benchmark set according to the product category sales cycle. Based on the pre-adjustment trigger identifier, the weight allocation of the price adjustment monitoring rule table is used to generate an operational control rule set. The operation control rule set is then used to conduct a competitor price-following response speed exceeding limit verification to determine the intervention threshold node. Based on the intervention threshold node, the operation control rule set is threshold-mapped to activate and output a price adjustment behavior sequence. The car owner click-but-not-purchase feature is extracted from the price adjustment behavior sequence to generate a recommendation gap coefficient. Based on the recommendation gap coefficient, complementary product association recommendation trend monitoring is performed on the price adjustment behavior sequence to generate sales compliance. Based on the sales compliance and the price adjustment behavior sequence, the intervention recurrence cycle is analyzed to generate an intervention recurrence cycle. Based on the intervention recurrence cycle and the operation control rule set, the deviation rate between search terms and transaction categories is calibrated to generate a category operation profile. The category operation profile is compared with the product operation data to generate an intelligent operation evaluation result.

2. The method according to claim 1, characterized in that, The step of coupling and parsing the product operation data and the competitor market data to generate a pre-adjustment trigger identifier includes: Based on the competitor market data, historical price series of competitors are extracted to construct a rigid rhythm distribution; A positional correlation table is generated by performing price rigidity cycle alignment matching between the rigidity rhythm distribution and the commodity operation data; The offset feature set is generated by identifying the price stagnation range of competitors that have not lowered their prices for a long time through the alignment association table. A pre-pricing trigger identifier is generated based on the cumulative trigger strength of the offset feature set.

3. The method according to claim 1, characterized in that, The step of performing inventory backlog rate detection and generating backlog rate distribution based on the product operation data includes: Based on the product operation data, identify the new product launch node records for each category and construct a digestion rate curve; The digestion rate curve is analyzed to obtain the slow-moving slope value by analyzing the section of rapid digestion rate drop. Based on the aforementioned slow-selling slope value, a new and old product category substitution association mapping is performed to obtain the slow-selling weight; Based on the sluggish sales weight and the sluggish sales slope value, backlog intensity mapping is performed to generate a backlog rate distribution.

4. The method according to claim 1, characterized in that, The step of establishing a price adjustment monitoring rule table by adapting the backlog rate distribution to the competitor benchmark set according to the product category sales cycle includes: The backlog rate distribution is divided into periodic rate groups according to the sales cycle of product categories; Differential benchmark matching is performed on the cycle rate group and the competitor benchmark set to generate a tolerance critical benchmark; The segmented price adjustment boundaries are set based on the aforementioned tolerance threshold. The segmented price adjustment boundaries are mapped and arranged to generate a price adjustment monitoring rule table.

5. The method according to claim 1, characterized in that, The step of generating an operational control rule set by weighting the price adjustment monitoring rule table based on the pre-adjustment trigger identifier includes: Analyze the distribution of the retention time in the favorites list based on the pre-price adjustment trigger identifier to locate the product category with a price adjustment gap; Based on the aforementioned price adjustment gap categories, weight configuration matching is performed to generate retention weight parameters; The retention weight parameters are sorted by priority gradient to determine the gap priority parameters; Based on the gap priority parameter and the price adjustment monitoring rule table, the rule priority is rearranged to generate an operation control rule set.

6. The method according to claim 1, characterized in that, The step of determining intervention threshold nodes by checking for excessive competitor price-following response speeds in the aforementioned operational control rule set includes: Based on the aforementioned operational control rule set, the price adjustment response delay of each price adjustment action is analyzed to form a price adjustment delay sequence; Delay threshold deviation is identified and a delay deviation feature set is generated for the price delay sequence; The intervention configuration table is obtained by filtering competitor actions whose response speed exceeds a critical threshold using the aforementioned time delay deviation feature set. The intervention configuration table is subjected to intervention intensity grading and extraction to generate intervention threshold nodes.

7. The method according to claim 1, characterized in that, The step of extracting the "car owner clicked but did not purchase" feature from the price adjustment behavior sequence to generate the recommendation gap coefficient includes: Based on the price adjustment behavior sequence, obtain the owner click records and transaction records for each product; The conversion gap between the car owner's click records and the transaction records is identified to generate a gap feature set; The gap category number group is determined by analyzing the gap category attributes using the gap feature set. Recommendation gap coefficients are generated by associating and arranging the gap category number group with the gap feature set.

8. The method according to claim 3, characterized in that, The step of analyzing the digestion rate curve for a sudden drop in digestion rate to obtain the slow-moving slope value includes: The digestion rate curve is used to identify replenishment nodes and locate sections where sales have not recovered. For the aforementioned sales failure to recover segment, identify the duration of sluggish sales and generate a sluggish sales feature set; The sluggish sales feature set is analyzed for time-varying changes to generate sluggish sales intensity parameters; Based on the aforementioned sluggish sales intensity parameter and the sales volume not yet recovered segment, a benchmark calibration is performed to generate a sluggish sales slope value.

9. The method according to claim 5, characterized in that, The step of analyzing the distribution of the time the product stays in the favorites list to locate the category with the price adjustment gap based on the pre-adjustment trigger identifier includes: The distribution characteristics of the retention time in the favorites are obtained by analyzing the distribution of retention time through the pre-price adjustment trigger identifier; Based on the aforementioned retention distribution characteristics, peak retention duration is identified to generate retention peak groups; A set of retention gap attributes is constructed by mapping the distribution characteristics of the retention peak group; Based on the aforementioned set of retained gap attributes, gap association matching is performed to locate the product category for price adjustment gaps.

10. An AI-based intelligent operation system for connected vehicle goods, characterized in that: include: The data acquisition module is used to collect product operation data and competitor market data, and to perform price rhythm coupling analysis on the product operation data and competitor market data to generate a pre-adjustment trigger identifier; The inventory detection module is used to perform inventory backlog rate detection on the product operation data to generate a backlog rate distribution, obtain a competitor benchmark set through the competitor market data, and establish a price adjustment monitoring rule table by adapting the backlog rate distribution to the competitor benchmark set according to the product category sales cycle. The rule allocation module is used to generate an operation control rule set by weighting the price adjustment monitoring rule table based on the pre-price adjustment trigger identifier, and to conduct a competitor price-following response speed exceeding limit verification to determine the intervention threshold node for the operation control rule set; The recommendation monitoring module is used to perform threshold mapping on the operation control rule set based on the intervention threshold node to activate and output the price adjustment behavior sequence, extract the car owner click-but-not-purchase feature for the price adjustment behavior sequence to generate a recommendation gap coefficient, and perform complementary product association recommendation trend monitoring on the price adjustment behavior sequence based on the recommendation gap coefficient to generate sales compliance. The results output module is used to perform intervention recurrence cycle analysis based on the sales compliance and the price adjustment behavior sequence to generate an intervention recurrence cycle, perform search term and transaction category deviation rate calibration based on the intervention recurrence cycle and the operation control rule set to generate a category operation profile, and compare the category operation profile with the product operation data to generate an intelligent operation evaluation result.

Images