AI intelligent marketing method for mining hot-selling competitive products based on category and price

Abstract

The application discloses an AI intelligent marketing hot-selling competitive product mining method based on category price, relates to the technical field of marketing, and comprises the following steps: establishing a unified time baseline and extracting the continuous sales curve of commodities under the target category, calculating the instantaneous slope of sales change in the sales curve, and determining the candidate area of short-period sales growth based on the mutation interval of the instantaneous slope; a causal factor set is constructed in the candidate area, and the channel expansion track and the promotion track are mapped to the sales curve point by point. The application generates a sales anomaly weight index through the joint constraint of causal residual error and user behavior density tensor, and writes back to the competitive product image, effectively filters false sales signals, and improves the accuracy of competitive product identification and marketing decision. At the same time, the short-time Fourier transform is introduced to decompose the sales curve, and the spectral domain attenuation coefficient is used to dynamically weaken the high-frequency anomaly, so as to ensure that the hot-selling weight truly reflects the long-term and periodic trend, thereby enhancing the reliability and analysis accuracy of the competitive product image.

Description

Technical Field

[0001] This invention relates to the field of marketing technology, specifically to an AI-powered intelligent marketing method for identifying best-selling competitors based on product category and price. Background Technology

[0002] In AI-powered intelligent marketing, identifying top-selling competitors based on product category and price range involves using big data and intelligent algorithms to conduct in-depth analysis of market sales across different price ranges within a specific product category. This process identifies competing products that boast high sales volume, rapid growth, positive user reviews, and potential market threats within the same segment. It goes beyond simply screening sales rankings; it combines category attributes, price range distribution, user preferences, and dynamic market trends to construct competitor profiles and uncover the most representative benchmarks for target users' actual purchasing choices. Through this approach, businesses can accurately pinpoint direct competitors, discover potential substitutes, and analyze their marketing strategies and product advantages. This provides data support for their own pricing, promotion, product selection, and differentiated positioning, enhancing the targeting and foresight of their marketing decisions.

[0003] The existing technology has the following shortcomings:

[0004] In existing technologies, some products may exhibit abnormal sales surges within a short period due to fraudulent order practices or sudden channel favoritism. However, such abnormal patterns are often not effectively identified and filtered, resulting in a false amplification effect of best-selling status during competitor profiling. Once this virtual inflation is used as the basis for subsequent analysis and strategy formulation, it can lead to biases in competitor identification, misjudging products that are not actually market leaders as core competitors. This, in turn, directly triggers incorrect adjustments to pricing strategies, resulting in serious consequences such as unbalanced resource allocation and delayed market response.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide an AI-powered intelligent marketing method for identifying best-selling competitors based on product category and price, in order to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an AI-powered intelligent marketing method for identifying top-selling competitors based on product category and price, comprising the following steps:

[0008] Establish a unified time baseline and extract continuous sales curves for products under the target category. Calculate the instantaneous slope of sales changes in the sales curves and determine candidate regions for short-cycle sales growth based on the abrupt change intervals of the instantaneous slopes.

[0009] A set of causal factors is constructed within the candidate region, and the channel expansion trajectory and promotion trajectory are mapped point by point to the sales curve. The causal residuals are calculated based on the mapping results to characterize the degree of deviation between sales growth and causal factors.

[0010] Based on the causal residual, a cross-channel consistency spectrum is generated, and the phase difference and synchronicity indicators of sales growth in each channel are projected onto the causal residual space. The projection results are then combined with the causal residual to form an abnormal signal.

[0011] Under the constraint of abnormal signals, a user behavior density tensor is constructed to model the density of the purchase time interval and geographical distribution of target users. The density peak is compared with the abnormal signals one by one to enhance the reliability of sales anomaly identification.

[0012] Under the joint constraints of user behavior density tensor and causal residual, an abnormal sales weight index is generated. Based on the abnormal sales weight index, false sales growth and normal sales trends are dynamically separated. The abnormal sales weight index is then written back to the competitor profile to complete the correction of the best-selling weight.

[0013] Driven by the abnormal sales weight index, the sales curve is transformed into a time spectrum signal. The sales changes are decomposed into low-frequency trends, mid-frequency fluctuations and high-frequency anomalies through short-time Fourier transform. The low-frequency trends and mid-frequency fluctuations are included in the hot-selling weight, and the high-frequency anomalies are dynamically weakened through the spectral domain attenuation coefficient to reduce the interference of false sales bursts on the competitor profile.

[0014] Preferably, the identification of short-term sales growth candidate regions based on sales data of products within the target category includes the following steps:

[0015] Establish a unified time baseline and extract continuous sales curves on this time baseline. Perform time mapping, interpolation, and moving average processing on sales data from different sources to construct complete and continuous sales curves.

[0016] The instantaneous slope of sales change is calculated on the continuous sales curve. The instantaneous slope curve is obtained through difference operation. The sales curve is smoothed before calculation to reduce noise interference, thereby identifying the abrupt change characteristics of the slope curve.

[0017] Candidate regions are determined based on the abrupt change points of the instantaneous slope curve. Abrupt change points are marked by threshold detection, and the time range before and after the abrupt change points is expanded to cover the entire growth process. If the interval between adjacent abrupt change points is insufficient, they are merged according to the set threshold. The starting sales, peak sales, ending sales and duration of the region are recorded.

[0018] Preferably, constructing a set of causal factors and calculating causal residuals within the candidate region includes the following steps:

[0019] External influencing factors related to sales changes are collected within the candidate region, and a set of causal factors is constructed, including channel expansion trajectory and promotion trajectory. The causal factors are then standardized and time-aligned to form a time series on a unified time baseline.

[0020] The causal factors are mapped point by point to the sales curve within the candidate area. At each time point, the values ​​of the causal factors are extracted and matched with the sales values. Weights are set according to historical contributions to establish the correspondence between causal factors and sales changes.

[0021] Based on the correspondence, causal residuals are calculated. Sales are predicted by causal factors and compared with actual sales to obtain residuals. The residuals are then superimposed with the abrupt change interval of sales slope in the time dimension to identify abnormal growth areas that cannot be explained by causal factors.

[0022] Preferably, generating a cross-channel consistency spectrum based on causal residuals includes the following steps:

[0023] The sales curves of each channel within the candidate region are extracted, and the local growth rate is calculated through a sliding window. The growth rate is converted into a phase signal, and the phase difference sequence is obtained with reference to the baseline curve.

[0024] Based on the phase difference sequence, the synchronicity index between channels is calculated. The cross-channel consistency matrix is ​​obtained through correlation analysis and coherence function. The consistency matrix is ​​then matched point by point with the causal residuals in the time dimension to generate a consistency spectrum.

[0025] The consistency spectrum is normalized and superimposed on the causal residual curve in the time dimension to form a joint anomaly signal. When the causal residual deviates positively and the consistency is low, the signal amplitude is enhanced to mark the abnormal growth area.

[0026] Preferably, when normalizing the consistency spectrum, the numerical range of the consistency spectrum is adjusted to be consistent with the numerical range of the causal residual, and a weighting coefficient is set during the superposition process, giving high weight to the causal residual and low weight to the consistency spectrum, so as to enhance the accuracy of abnormal signal identification.

[0027] Preferably, constructing a dense tensor of user behavior under the constraint of anomalous signals includes the following steps:

[0028] Under the constraint of abnormal signals, the user's order time is extracted from the transaction data of the candidate region, the time interval of consecutive orders is calculated, and the time interval density distribution is formed by kernel density estimation;

[0029] Extract the geographic location information corresponding to the transaction data, map the order location to a two-dimensional space, and obtain the geographic density distribution through density clustering;

[0030] By combining the time interval density distribution with the geographical density distribution, a user behavior density tensor is constructed. The density peaks are then compared with abnormal signals one by one in the tensor space to enhance the reliability of sales anomaly identification.

[0031] Preferably, generating the sales anomaly weight index under the joint constraints of the user behavior density tensor and causal residuals includes the following steps:

[0032] Feature vectors are extracted based on dense tensors of user behavior and causal residuals, a set of weight factors is constructed, and they are projected onto a unified numerical space through feature standardization.

[0033] An abnormal sales weight index is generated based on the set of weighting factors. Abnormal signals are amplified by weighted superposition and exponential function. The output results are normalized to limit the range of values.

[0034] The abnormal sales weight index is used to dynamically peel off the sales curves of candidate regions, weakening the abnormal growth part with a decay coefficient and retaining the normal trend part.

[0035] The abnormal sales weight index is written back to the competitor profile to correct the original best-selling weight, so as to avoid the bias in competitor identification caused by the inflation of false sales.

[0036] Preferably, converting the sales curve into a time-spectrum signal under the drive of the sales anomaly weighting index includes the following steps:

[0037] Driven by the abnormal sales weight index, the sales curves of candidate regions are weighted and adjusted, and divided into adjacent time windows to form the input signal for spectrum analysis.

[0038] Applying short-time Fourier transform to the time window yields the spectral distribution of low-frequency trends, mid-frequency fluctuations, and high-frequency anomalies.

[0039] Low-frequency trends and mid-frequency fluctuations are directly incorporated into the sales weighting, and the sales anomaly weighting index is used to attenuate the amplitude in the high-frequency anomaly range in order to weaken the abnormal signal.

[0040] The low-frequency trend and mid-frequency fluctuation are merged and updated to the best-selling weights in the competitor profile, and the attenuated high-frequency anomalies are used as monitoring indicators and written back to the competitor profile.

[0041] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0042] This invention constructs a joint constraint mechanism based on causal residuals and user behavior density tensors, which can accurately identify abnormal sales growth caused by fraudulent orders or channel bias within a short period, effectively filtering out false sales signals. By dynamically calculating the abnormal sales weight index and writing it back to the competitor profile, it effectively avoids the amplification effect of false sales on the best-selling weight, thereby significantly improving the accuracy of competitor identification and the reliability of marketing decisions, and solving the problem that competitor profiles are easily distorted by abnormal data in existing technologies.

[0043] This invention introduces a short-time Fourier transform to decompose the sales curve into three types of signals: low-frequency trends, mid-frequency fluctuations, and high-frequency anomalies. It then uses a spectral attenuation coefficient to dynamically weaken high-frequency anomalies, effectively reducing the interference of abnormal sales in the frequency domain. This approach not only enhances the ability to identify true market trends but also allows the best-selling weights to more accurately reflect a product's competitiveness in long-term and cyclical dimensions, comprehensively improving the adaptability and analytical accuracy of competitor profiling under multi-frequency behavioral patterns. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0045] Figure 1 This is a flowchart illustrating the AI-powered intelligent marketing method for identifying best-selling competitors based on product category and price, as described in this invention. Detailed Implementation

[0046] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0047] This invention provides, for example Figure 1 The AI-powered intelligent marketing method shown here, based on product category and price range, for identifying top-selling competitors, includes the following steps:

[0048] S101. Establish a unified time baseline and extract the continuous sales curve of products under the target category. Calculate the instantaneous slope of sales changes in the sales curve and determine the candidate region for short-cycle sales growth based on the abrupt change interval of the instantaneous slope.

[0049] Identifying candidate regions for short-term sales growth based on sales data for products within the target category includes the following steps:

[0050] First, a unified time baseline is established, and a continuous sales curve is extracted from this baseline. To do this, sales data from different sources are imported into the same time series structure, mapping all sales data to a continuous time axis measured in minutes, hours, or days. The timestamps from different sources are formatted to ensure uniform time precision. Then, missing sales data points on the time axis are filled in using piecewise linear interpolation, ensuring that each time point on the time axis corresponds to a sales value. For channel data with overly dense sampling, a sliding window averaging method is used to downsample the data, aligning it with the time interval of the unified time baseline. Next, sales data from all channels are summed or kept independent for each channel to construct a continuous sales curve under the unified time baseline. This sales curve fully covers the sales trajectory of the target product category within a given time period, without any breakpoints, jumps, or abnormal fluctuations caused by inconsistent sampling granularity. This continuous sales curve ensures that subsequent steps are performed within a unified time reference framework, avoiding incomparable results due to differences in data sources.

[0051] After obtaining the continuous sales curve, the instantaneous slope of sales changes is calculated to capture the dynamic growth characteristics of sales over time. To do this, the continuous sales curve is differenced between two adjacent time points to obtain the change in sales, which is then divided by the time interval between the two adjacent time points to obtain the instantaneous slope for that interval. To improve the stability and noise resistance of the calculation, the continuous sales curve can be first Gaussian smoothed to reduce outliers in the instantaneous slope caused by random fluctuations. During the calculation, the instantaneous slope at each time point is saved as a new time series, forming a slope curve that corresponds one-to-one with the sales curve. When the sales curve experiences a sharp rise within a short period, the corresponding instantaneous slope value will show a significant positive peak, and the difference between the slope and the adjacent time point will widen significantly. Mathematically, this phenomenon is represented by a sudden change in the slope curve, indicating an abnormally rapid increase in the sales growth rate. By calculating the instantaneous slope point by point and constructing slope curves, key moments in the sales curve that may represent abnormal growth can be clearly identified.

[0052] After obtaining the instantaneous slope curve, candidate regions for short-cycle sales growth are constructed based on abrupt change points. Specifically, the instantaneous slope curve is first scanned to detect points exceeding a preset threshold. This threshold can be calculated based on the long-term average slope and standard deviation of historical sales curves, for example, set as the average plus three times the standard deviation. When the instantaneous slope at a certain time point is greater than this threshold, that time point is marked as a candidate abrupt change point. Next, the candidate abrupt change points are extended forward and backward by a certain time range, for example, by 2 to 4 sampling intervals each, to cover the initial and declining phases of sales growth. If the interval between adjacent abrupt change points is less than a set time threshold, these abrupt change points are merged into a continuous candidate region to avoid incorrectly splitting the same growth process into multiple regions. For each candidate region, the initial sales, peak sales, and ending sales of the sales curve within that region, as well as the corresponding duration and average slope, need to be recorded simultaneously for use in subsequent causal factor matching and anomaly identification. The resulting candidate regions are not just single instantaneous slope abrupt changes, but cover the entire process of sales from a stable state to a sharp increase and then back to stability. This ensures that subsequent steps can be analyzed based on the complete growth trajectory without missing important contextual information.

[0053] By implementing the above steps, a unified time baseline is first established to ensure the comparability and continuity of data over time. Secondly, the dynamic characteristics of sales change rates are captured through point-by-point calculation of instantaneous slopes. Finally, continuous candidate regions are formed through abrupt change point expansion and regional aggregation. This process enables the accurate identification of short-cycle abnormal growth in the sales curve. This method not only avoids misjudgments of false signals caused by short-term promotions or fraudulent orders in existing technologies, but also ensures the integrity and robustness of candidate region identification.

[0054] S102. Construct a set of causal factors within the candidate region, map the channel expansion trajectory and promotion trajectory point by point to the sales curve, and calculate the causal residual based on the mapping results to characterize the degree of deviation between sales growth and causal factors.

[0055] To provide a causal explanation for the authenticity of sales growth within candidate regions, a set of causal factors is constructed within these regions. Then, by mapping channel expansion trajectories and promotional trajectories point-by-point to the sales curve, causal residuals are calculated to characterize the degree of deviation between sales growth and the causal factors. This process includes the following steps:

[0056] Within the identified candidate regions, external influencing factors related to sales changes are collected, and a causal factor set is constructed. The causal factor set primarily includes two types of data: channel expansion trajectory and promotional trajectory. Channel expansion trajectory refers to changes in the target product's sales channels within the candidate region's time frame, such as whether new online e-commerce channels have been added, whether new distribution outlets have been opened, or whether platform recommendation exposure has increased. Promotional trajectory refers to various promotional activities implemented for the target product within the candidate region's time frame, such as discounts, coupons, flash sales, live streaming promotions, and cross-category joint promotions. To ensure the causal factor set comprehensively covers external variables related to sales, standardization and time alignment operations are performed on different data sources, ensuring all causal factors are expressed on a time baseline completely consistent with the candidate regions. Each causal factor is represented as a time series, where each time point corresponds to a numerical value reflecting the factor's strength at that moment. For example, whether a new distribution node was added at a certain time in the channel expansion trajectory can be represented by a binary number, and promotional intensity can be represented by a discount percentage or promotional budget size.

[0057] The aforementioned causal factors are mapped point-by-point to the sales curve within the candidate region to establish a correspondence between causal factors and sales changes. Specifically, at each time point within the candidate region, the values ​​of all causal factors corresponding to that time point need to be extracted and paired with the sales figures at that time point to form a time-point-level mapping between causal factors and sales. To ensure the accuracy of the mapping, the importance of different causal factors can be considered using a weighted approach. For example, channel expansion factors and promotion factors can be assigned different weights, which can be determined based on the statistical analysis results of the contribution of causal factors to sales in historical data. Through this point-by-point mapping operation, it can be determined whether fluctuations on the sales curve in each time segment of the candidate region correspond to a certain channel expansion event or promotional activity. If the sales change and the causal factor change occur synchronously, it indicates that the sales growth has a causal explanation; if the sales change suddenly increases when the causal factors remain unchanged, it means that there are additional anomalies.

[0058] After mapping causal factors to the sales curve point by point, it is necessary to calculate the causal residual based on the mapping results to measure the degree of deviation between sales growth and the causal factors. Specifically, a sales prediction function based on causal factors is established. This function uses the values ​​of channel expansion trajectory and promotion trajectory to predict the theoretically expected sales increase at each time point, and then compares the predicted sales value with the actual sales value. The difference between the two is the causal residual, and the magnitude of the residual directly reflects the part of sales change that cannot be explained by known causal factors. When the residual shows a significant positive deviation at consecutive time points, it means that the sales growth is much higher than the level that promotional activities and channel expansion can explain, which is very likely to be due to interference from fraudulent order behavior or abnormal channel traffic tilt. When the residual shows a negative deviation at multiple time points, it may indicate that the promotion or channel expansion has not achieved the expected effect, and the sales have not grown as theoretically predicted. By calculating the causal residual, we can not only identify which parts of the sales curve belong to the explainable normal growth and which parts belong to the unexplained abnormal growth, but also provide a basis for subsequent amplification of abnormal signals and multi-dimensional cross-validation.

[0059] Causal residuals are used as a preliminary measure of anomaly signals in candidate regions, and then combined with sales curves for enhanced analysis. This process involves comparing the causal residuals and sales curves point-by-point over time, superimposing areas of sudden increases in the residuals with instantaneous slope abrupt changes in the sales curve. If the causal residuals deviate continuously and significantly within the sales slope abrupt change range, the candidate region can be marked as an anomalous growth region, indicating that the sales growth cannot be explained by normal channels or promotions, exhibiting obvious characteristics of spurious inflation. Simultaneously, statistical characteristics of the residuals, such as mean, variance, kurtosis, and skewness, can be calculated for each candidate region to further quantify the strength of the anomaly signal. These statistical characteristics will continue to be used in subsequent cross-channel consistency spectrum construction and user behavior density tensor modeling, forming a progressively validating chain of anomaly detection. In this way, causal residuals are not merely a single measure of deviation, but become a core indicator for analyzing the authenticity of sales within candidate regions.

[0060] In summary, through the above steps, a complete set of causal factors within the candidate region is constructed, and the channel expansion trajectory and promotion trajectory can be mapped point by point to the sales curve. Furthermore, quantitative characterization of the abnormal portion of sales growth is achieved through causal residual calculation. This implementation method, through residual analysis, establishes a mechanism for distinguishing between the explainable and non-explainable portions of sales growth, laying a data and methodological foundation for subsequent amplification of abnormal signals, cross-channel consistency analysis, and user behavior verification.

[0061] S103. Generate a cross-channel consistency spectrum based on causal residuals, project the phase difference and synchronicity indicators of sales growth in each channel onto the causal residual space, and jointly superimpose the projection results with the causal residuals to form an abnormal signal.

[0062] To more accurately identify anomalous components in sales growth that cannot be explained by causal factors, cross-channel consistency analysis is introduced based on causal residuals. By generating a cross-channel consistency spectrum, the phase difference and synchronicity indicators of sales growth across channels are projected onto the causal residual space and jointly superimposed with the causal residuals to form clear anomalous signals. This process includes the following steps:

[0063] The sales curves of each sales channel within the candidate region are extracted independently, and the phase characteristics of their growth trajectories are calculated for integration with causal residuals. In practice, for each channel's sales curve, a local growth rate curve is first calculated using a sliding window approach, and then this growth rate curve is converted into a phase signal in time series form. The phase signal refers to the deviation of a channel's sales growth fluctuations on the time axis relative to a baseline time point during the sales growth process. To ensure the comparability of phase characteristics across channels, a unified reference curve is selected as the benchmark. For example, a weighted average curve of sales from all channels or the most representative channel curve from long-term historical data can be chosen as the benchmark. By calculating the time-dimensional offset between each channel curve and the benchmark curve, a phase difference sequence can be obtained. The phase difference sequence reveals the leading or lagging relationship of sales growth in different channels over time and is an important basis for determining whether there is synchronicity between channels.

[0064] After obtaining the phase difference sequences for each channel, the synchronicity indices between channels are further calculated, and these indices are correlated with causal residuals. Specifically, firstly, the correlation coefficients between sales curves of different channels are calculated using cross-correlation analysis, and the correlation strength at different time intervals within the candidate region is extracted. Higher correlation strength indicates more synchronized sales changes across different channels within that time interval; lower correlation strength indicates significant differences in sales growth performance across different channels. To avoid the limitations of a single correlation indicator, a coherence function can be introduced to calculate the synchronicity of sales curves across different channels in the frequency domain, obtaining a cross-channel synchronicity spectrum distribution. Subsequently, these synchronicity indices are combined with the phase difference sequences obtained in the previous step to form a cross-channel consistency matrix. Each element of this consistency matrix represents the phase difference and synchronicity level between a pair of channels at a certain point in time. Next, this consistency matrix is ​​projected onto the causal residual space, that is, each element in the matrix is ​​matched point-to-point with the causal residual at the same point in time, and a consistency spectrum is generated through vectorized superposition. In this process, if the causal residual value is large and the cross-channel consistency is low, it indicates that the sales growth lacks consistent support across multiple channels and is likely caused by abnormal behavior in a single channel. Conversely, if the causal residual value is small and the cross-channel consistency is high, it indicates that the sales growth is supported by multiple channels and is a relatively normal market phenomenon.

[0065] After generating the cross-channel consistency spectrum, it is jointly superimposed with the causal residuals to form a clearer anomalous signal. In the specific implementation, the consistency spectrum is first normalized to ensure its numerical range matches that of the causal residuals, thus avoiding deviations caused by different units. Subsequently, the normalized cross-channel consistency spectrum and the causal residual curve are superimposed point-by-point over time to obtain the joint anomalous signal curve. On this curve, when the causal residuals exhibit a persistent positive deviation, and the consistency spectrum shows asynchrony or excessive phase difference between channels, the amplitude of the joint signal will significantly increase, forming prominent anomalous peaks. These anomalous peaks clearly mark unreasonable parts of sales growth within the candidate region. Even if this growth is explosive within a single channel, it will still be identified as anomaly in the joint signal due to the lack of cross-channel synchronization support. To further improve the robustness of the anomalous signal, weighting coefficients can be introduced during the superposition process, for example, assigning a higher weight to the causal residuals and a relatively lower weight to the consistency spectrum, to emphasize the impact of insufficient causal explanation. The resulting joint anomaly signal not only retains the anomaly explanation capability of causal residuals, but also introduces cross-channel consistency as an auxiliary verification method, thereby significantly reducing the risk of misjudgment caused by occasional promotions or local traffic tilt.

[0066] Through the progressive steps described above, phase features are first extracted from channel sales curves to ensure the quantification of temporal differences between different channels. Secondly, synchronicity indicators are calculated and projected into the causal residual space, achieving an organic combination of causal explanation and cross-channel consistency. Finally, anomaly signals are generated through joint superposition, making the identification of abnormal sales growth more intuitive and accurate. This specific implementation method is the first to combine causal residuals with cross-channel consistency analysis, considering both whether sales growth can be explained by external factors and the coordination of different channels during the growth process, thus significantly improving the reliability and accuracy of anomaly identification under dual constraints. This method can not only address fraudulent order-boosting behavior in a single channel but also identify false surges caused by channel traffic imbalances.

[0067] S104. Construct a user behavior density tensor under the constraint of abnormal signals, model the density of the purchase time interval and geographical distribution of target users, and compare the density peak with the abnormal signals one by one to enhance the reliability of sales anomaly identification.

[0068] To further improve the reliability of sales anomaly identification, user behavior is introduced for cross-validation under the constraint of generated anomaly signals. By constructing a user behavior density tensor, the behavioral characteristics of target users in terms of purchase time interval and geographical distribution are transformed into a density model. The density peaks are then compared with the anomaly signals one by one, thereby achieving multi-dimensional anomaly identification. This process includes the following steps:

[0069] Under the constraint of the obtained anomaly signals, the purchase behavior information of target users is extracted from the transaction data within the candidate region, and a dense distribution of purchase time intervals is constructed based on this information. Specifically, for all valid orders within the candidate region, the unique identifier of the ordering user and the corresponding order time are extracted one by one. Then, the time interval between two consecutive orders is calculated on a user-by-user basis, forming a time interval sequence. Under normal circumstances, the time interval distribution of different users has strong randomness and diversity, and the dense distribution exhibits a relatively smooth shape. However, under anomaly conditions, the time interval distribution shows a significant concentration phenomenon, such as a large number of orders being completed by a small number of users in a short period of time, or a large number of orders repeating within a very short time interval. To characterize this concentration feature, the time interval sequence is mapped to a unified time interval coordinate axis, and the probability density function is calculated on this axis using the kernel density estimation method, thus forming a dense distribution of time intervals. This dense distribution can intuitively reflect the concentration and anomaly of user purchase behavior in the time dimension and is an important component for the subsequent construction of the dense density tensor.

[0070] While obtaining the density distribution of purchase time intervals, the density model of the purchase distribution of target users in geographic space is also performed. Specifically, for all orders within the candidate region, the geographic location information corresponding to the orders is extracted, such as the latitude and longitude coordinates of the delivery address, and these coordinate points are mapped to a two-dimensional geographic space. Under normal circumstances, the geographic distribution should match the market coverage of the target product category, showing an even distribution across multiple cities and regions; however, in abnormal situations, orders are often concentrated in a very small number of geographic locations, or even within the same city or neighborhood. To characterize this abnormal concentration, geographic coordinate points need to be input into a spatial clustering algorithm, such as a density-based clustering method, to identify areas with highly concentrated spatial distribution and calculate the order density within each cluster. Furthermore, a geographic density distribution map can be constructed on a two-dimensional plane to visually represent the concentration of purchasing behavior. In this way, if sales growth within a candidate region is found to be accompanied by abnormal geographic concentration, it can corroborate the anomaly signal, thereby improving the reliability of anomaly identification.

[0071] After obtaining the purchase time interval density distribution and geographical density distribution, these two are combined to construct a user behavior density tensor. This tensor is then compared one by one with abnormal signals to enhance the reliability of the identification. Specifically, the user behavior density tensor is represented in three dimensions: one dimension is time interval density, another is geographical density, and the third is the time series. By modeling user behavior within candidate regions in the three-dimensional tensor space, concentrated features in both the time and spatial dimensions can be captured simultaneously, and their evolution trend over time can be observed. In this tensor, if peak regions with excessively dense time intervals and geographically concentrated distributions appear simultaneously within a certain time period, and these peak regions highly overlap with abnormal signals in time, it indicates that the sales growth within that time period is highly likely to be caused by abnormal transaction behavior. To ensure the rigor of the matching, the intensity of the density peaks needs to be calculated in the tensor space, and a correlation analysis is performed with the amplitude of the abnormal signals. If the correlation coefficient between the two is significantly higher than a set threshold, the sales growth is confirmed to be an abnormal signal; if the correlation is low, it indicates that the abnormal signal may be caused by other unobserved factors, requiring further verification in subsequent stages. By comparing each other one by one, the dense tensor of user behavior not only serves as an auxiliary verification method, but also as an amplifier of the intensity of abnormal signals. This makes anomaly identification no longer rely solely on sales curves and causal factor analysis, but integrates user-level behavioral characteristics, thereby achieving multi-dimensional cross-validation.

[0072] By organically combining the above steps, the abnormal concentration of user order behavior in the time dimension was first identified through the purchase time interval distribution. Secondly, the abnormal clustering of orders in the spatial dimension was revealed through the geographical density distribution. Finally, by constructing a user behavior density tensor and comparing it one by one with abnormal signals, joint verification in both time and space dimensions was achieved. This specific implementation couples the time interval characteristics and geographical distribution characteristics of user behavior into the same density tensor and performs superimposed analysis with abnormal signals under causal residual constraints, thus forming a multi-level verification framework across sales volume and user behavior dimensions. This framework can not only effectively exclude normal growth caused by promotional activities or channel adjustments but also accurately identify false surges caused by fraudulent orders or abnormal channel traffic.

[0073] S105. Under the joint constraints of the user behavior density tensor and causal residual, an abnormal sales weight index is generated. Based on the abnormal sales weight index, false sales growth and normal sales trends are dynamically separated. The abnormal sales weight index is written back to the competitor profile to complete the correction of the best-selling weight.

[0074] To isolate fraudulent sales growth and correct the best-selling weights in competitor profiles, an abnormal sales weight index is generated under the joint constraints of user behavior density tensor and causal residuals. This index is then used to distinguish between fraudulent sales growth and normal sales trends. Finally, the abnormal sales weight index is written back into the competitor profile. This process includes the following steps:

[0075] Under the constraints of the obtained user behavior density tensor and causal residuals, a set of weighting factors for measuring the authenticity of sales is constructed. Specifically, the user behavior density tensor provides a concentrated indicator of user ordering behavior across time intervals and geographical distribution dimensions, while the causal residuals reflect the portion of sales growth that cannot be explained by channel expansion and promotions. In this embodiment, the peak intensity, peak duration, and frequency of peak occurrence in the user behavior density tensor are extracted as a set of behavioral feature vectors, while the mean, variance, and abnormal deviation magnitude in the causal residual sequence are extracted as a set of residual feature vectors. Subsequently, the behavioral feature vectors and residual feature vectors are projected onto the same numerical space using feature standardization methods, enabling comprehensive calculation on the same dimension. The core of this process lies in providing quantitative input factors for the subsequent generation of the sales anomaly weight index through joint modeling of user behavior and causal residuals.

[0076] After obtaining the set of weighting factors, a sales anomaly weighting index is generated based on this set. Specifically, an exponential function is first defined to weight and superimpose user behavior features and residual features according to preset weight parameters, amplifying the contribution of abnormal signals through exponentialization. For example, when the peak of user behavior density and the peak of causal residuals highly overlap in time, their superposition result will be non-linearly amplified by the exponential function, resulting in a significantly high value in the sales anomaly weighting index. To ensure the stability of the calculation results, the output range of the exponential function can be normalized, so that the value of the sales anomaly weighting index is distributed between 0 and 1. Values ​​close to 0 indicate that sales growth can be interpreted as a normal trend, while values ​​close to 1 indicate that sales growth is highly likely to be spurious inflation. In this way, the sales anomaly weighting index achieves a unified expression of complex multidimensional data, allowing anomaly identification results to be presented in an intuitive and comparable numerical form.

[0077] After generating the abnormal sales weight index, this index is used to dynamically peel off the sales curves within the candidate region, thereby distinguishing between false growth and normal trends. Specifically, the original sales curves in the candidate region are decomposed into two parts: one part represents the normal sales trend, and the other part represents abnormal sales growth. During the calculation process, for time points where the abnormal sales weight index is higher than a set threshold, the corresponding sales increment is classified as the abnormal growth part, and its contribution is weakened by a decay coefficient; while for time points where the abnormal sales weight index is lower than the threshold, the sales increment is completely retained as part of the normal trend. Furthermore, a sliding window method can be used to smooth the abnormal sales weight index to avoid excessive impact of fluctuations at a single time point on the overall peeling result. Ultimately, the sales curve after dynamic peeling accurately reflects the true sales growth trend of the target product within the candidate region, while the peeled-off abnormal part clearly identifies possible order-brushing behavior or false inflation caused by channel traffic skew. This process not only improves the authenticity of sales data but also provides a reliable basis for subsequent competitor profiling.

[0078] After stripping away any false sales growth, the abnormal sales weight index is written back into the competitor profile to correct the best-selling weights. Specifically, the competitor profile is a multi-dimensional market performance representation tool, with best-selling weights being one of the core indicators. In this implementation, the abnormal sales weight index is used as a correction factor to dynamically adjust the original best-selling weights. For example, if a product in the original competitor profile gains a high best-selling weight due to a short-term sales surge, and its corresponding abnormal sales weight index is close to 1, the product's best-selling weight is lowered to a reasonable range using an index correction formula, thus avoiding weight inflation caused by false sales. Simultaneously, for products with low abnormal sales weight indices, their best-selling weights can remain unchanged or be appropriately increased to highlight their true market competitiveness. By writing the corrected best-selling weights back into the competitor profile, the final competitor profile more accurately and comprehensively reflects the market landscape, helping companies make more precise decisions regarding competitor identification, pricing strategies, and resource allocation.

[0079] Through the above steps, firstly, feature vectors of user behavior density tensors and causal residuals were extracted and a set of weighting factors was established. Secondly, a sales anomaly weight index, which uniformly expresses the degree of anomaly, was generated based on this set. Thirdly, the sales curves within the candidate region were dynamically stripped using this index to distinguish between false growth and normal trends. Finally, the index was written back to the competitor profile to correct the best-selling weights. This specific implementation not only establishes a joint constraint mechanism across behavioral and causal dimensions but also achieves quantitative differentiation and correction of false sales through indexation and dynamic stripping methods, thereby significantly improving the accuracy of competitor profiles.

[0080] S106. Driven by the abnormal sales weight index, the sales curve is converted into a time spectrum signal. The sales change is decomposed into low-frequency trend, mid-frequency fluctuation and high-frequency anomaly through short-time Fourier transform. The low-frequency trend and mid-frequency fluctuation are included in the hot-selling weight, and the high-frequency anomaly is dynamically weakened through the spectral domain attenuation coefficient to reduce the interference of false sales bursts on the competitor profile.

[0081] To completely reduce the interference of false sales surges on competitor profiling, driven by a sales anomaly weighting index, the sales curve is transformed into a time-spectrum signal. Short-time Fourier transform is then used to decompose sales changes into low-frequency trends, mid-frequency fluctuations, and high-frequency anomalies. Finally, a spectral attenuation coefficient is used to dynamically weaken the high-frequency anomaly component, thereby achieving precise correction of the best-selling weights. This process includes the following steps:

[0082] Driven by the obtained sales anomaly weighting index, the sales curves of the candidate regions are transformed into time signals suitable for spectral analysis. Specifically, the sales anomaly weighting index, as a dynamic factor, weights and adjusts each time point in the sales curve, appropriately weakening the sales curve value at time points with higher anomaly weights, while maintaining the original amplitude at time points with lower anomaly weights. In this way, the transformed time signal already embeds preliminary anomaly suppression information in the time dimension before entering the frequency domain analysis. Next, the weighted sales curve is discretized, dividing it into multiple adjacent time windows. The length of each time window is set according to the time span of the candidate region, for example, 5 days, 7 days, or 10 days can be selected as a window. Each window will serve as the input unit for the short-time Fourier transform to ensure that the dynamic characteristics of the sales curve at different time scales can be captured. The core of this process is that, guided by the sales anomaly weighting index, the time signal processed in the spectral analysis stage already possesses the prior constraints to distinguish between normal trends and abnormal fluctuations.

[0083] After obtaining the weighted time signal, a short-time Fourier transform (SFT) is applied to each time window, transforming the sales curve in the time domain into a spectral distribution in the frequency domain. The specific steps of the SFT include: first, selecting an appropriate window function, such as a Hamming window or a Gaussian window, to avoid the influence of boundary effects on the transformation result; then, windowing the sales signal within each time window and calculating its amplitude and phase at different frequency components to obtain the spectrum for that time window. By performing the SFT on all time windows, the characteristics of sales changes can be observed simultaneously in both the time and frequency dimensions. In the spectrum, the low-frequency component corresponds to the long-term trend of sales, such as stable growth over several weeks or months; the mid-frequency component corresponds to periodic fluctuations, such as sales fluctuations caused by weekly promotions or holiday promotions; and the high-frequency component corresponds to sharp fluctuations in a short period, often related to fraudulent order practices, sudden traffic spikes, or abnormal promotional activities. Therefore, the SFT not only reveals the overall trend of the sales curve but also finely distinguishes growth patterns within different frequency ranges, providing a clear decomposition basis for subsequent anomaly reduction processing.

[0084] After completing the spectral decomposition of the sales curve, low-frequency trends, mid-frequency fluctuations, and high-frequency anomalies are categorized and processed to play different roles in the adjustment of hot-selling weights. Specifically, low-frequency trends are considered the core part of sales growth and are directly included in the calculation of hot-selling weights; mid-frequency fluctuations are considered as regular market fluctuations and should also be retained and included in the hot-selling weights to ensure that the competitor profile accurately reflects the impact of cyclical factors on the market; and for high-frequency anomalies, further weakening processing is required in conjunction with the sales anomaly weight index. In this implementation, the frequency range of high-frequency components is first identified in the spectrum, and their corresponding amplitude distribution is extracted. Then, the sales anomaly weight index is used as an adjustment factor to attenuate these amplitudes. Specifically, the sales anomaly weight index is multiplied point-by-point with the high-frequency amplitude to form a spectral adjustment vector. In time windows with high anomaly weights, high-frequency amplitudes are significantly weakened, while in time windows with low anomaly weights, high-frequency amplitudes remain relatively intact. This spectral adjustment based on the sales anomaly weight index allows high-frequency anomalies to be dynamically weakened without affecting the authenticity of low-frequency and mid-frequency components.

[0085] After dynamically weakening high-frequency anomalies, low-frequency trends and mid-frequency fluctuations are merged back into the best-selling weights, and the processed spectrum results are written back into the competitor profile to complete the final correction. Specifically, the amplitude components of low-frequency and mid-frequency data are re-integrated to form the corrected best-selling weight values, which are then updated in the corresponding best-selling dimension indicators in the competitor profile. Simultaneously, the weakened high-frequency anomalies are retained as an additional monitoring indicator to flag potential risks of false sales spikes. In this way, the corrected competitor profile not only accurately reflects the long-term trends and cyclical fluctuations of products in the market but also provides timely warnings when anomalies occur, preventing companies from misjudging false sales as genuine competitiveness. The final competitor profile has a dual advantage: firstly, it is more accurate and reliable in terms of best-selling weights, avoiding interference from false spikes; secondly, it is more sensitive in risk monitoring, providing forward-looking support for companies to formulate pricing, promotion, and product selection strategies.

[0086] Through the implementation of the above steps, the sales curve is first weighted and divided into time windows under the drive of the abnormal sales weight index. Secondly, time spectrum decomposition is achieved through short-time Fourier transform. Thirdly, low-frequency, mid-frequency and high-frequency are effectively distinguished through classification and anomaly reduction. Finally, the correction results are written back to the competitor profile to complete the adjustment of the best-selling weight. This not only removes false growth in the time domain, but also dynamically reduces high-frequency anomalies in the frequency domain through the spectral attenuation coefficient, thereby constructing a two-dimensional correction mechanism across the time and frequency domains, which greatly improves the authenticity and reliability of the competitor profile.

[0087] This invention constructs a joint constraint mechanism based on causal residuals and user behavior density tensors, which can accurately identify abnormal sales growth caused by fraudulent orders or channel bias within a short period, effectively filtering out false sales signals. By dynamically calculating the abnormal sales weight index and writing it back to the competitor profile, this invention effectively avoids the amplification effect of false sales on the best-selling weight, thereby significantly improving the accuracy of competitor identification and the reliability of marketing decisions, and solving the problem that competitor profiles are easily distorted by abnormal data in existing technologies.

[0088] This invention further introduces short-time Fourier transform to decompose the sales curve into three types of signals: low-frequency trends, mid-frequency fluctuations, and high-frequency anomalies. It also incorporates a spectral attenuation coefficient to dynamically weaken high-frequency anomalies, effectively reducing the interference of abnormal sales in the frequency domain. This approach not only enhances the ability to identify true market trends but also allows the best-selling weights to more accurately reflect a product's competitiveness in long-term and cyclical dimensions, comprehensively improving the adaptability and analytical accuracy of competitor profiling under multi-frequency behavioral patterns.

[0089] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. An AI-powered intelligent marketing method for identifying top-selling competitors based on product category and price range, characterized by: Includes the following steps: Establish a unified time baseline and extract continuous sales curves for products under the target category. Calculate the instantaneous slope of sales changes in the sales curves and determine candidate regions for short-cycle sales growth based on the abrupt change intervals of the instantaneous slopes. A set of causal factors is constructed within the candidate region, and the channel expansion trajectory and promotion trajectory are mapped point by point to the sales curve. The causal residuals are then calculated based on the mapping results. Based on the causal residual, a cross-channel consistency spectrum is generated. The phase difference and synchronicity indicators of sales growth in each channel are projected onto the causal residual space, and the projection results are jointly superimposed with the causal residual to form an abnormal signal. Under the constraint of anomalous signals, a user behavior density tensor is constructed to model the density of the purchase time interval and geographical distribution of target users, and the density peak is compared with the anomalous signals one by one. Under the joint constraints of user behavior density tensor and causal residual, an abnormal sales weight index is generated. Based on the abnormal sales weight index, false sales growth and normal sales trends are dynamically separated. The abnormal sales weight index is written back to the competitor profile to correct the best-selling weight. Driven by the sales anomaly weighting index, the sales curve is transformed into a time spectrum signal. The sales changes are decomposed into low-frequency trends, mid-frequency fluctuations and high-frequency anomalies through short-time Fourier transform. The low-frequency trends and mid-frequency fluctuations are included in the hot-selling weights, and the high-frequency anomalies are dynamically weakened through the spectral attenuation coefficient.

2. The AI-powered intelligent marketing method for identifying best-selling competitors based on product category and price, as described in claim 1, is characterized in that... Identifying short-term sales growth candidate regions based on sales data for products within a target category includes the following steps: Establish a unified time baseline and extract continuous sales curves on this time baseline. Perform time mapping, interpolation, and moving average processing on sales data from different sources to construct complete and continuous sales curves. The instantaneous slope of sales change is calculated on the continuous sales curve. The instantaneous slope curve is obtained through difference operation. The sales curve is smoothed before calculation, and the abrupt change characteristics of the slope curve are identified. Candidate regions are determined based on the abrupt change points of the instantaneous slope curve. Abrupt change points are marked by threshold detection, and the time range before and after the abrupt change points is expanded to cover the entire growth process. If the interval between adjacent abrupt change points is insufficient, a threshold is set and the regions are merged.

3. The AI-powered intelligent marketing method for identifying top-selling competitors based on product category and price, as described in claim 1, is characterized in that... Constructing a set of causal factors and calculating causal residuals within candidate regions includes the following steps: External influencing factors related to sales changes are collected within the candidate region, and a set of causal factors is constructed, including channel expansion trajectory and promotion trajectory. The causal factors are then standardized and time-aligned to form a time series on a unified time baseline. The causal factors are mapped point by point to the sales curve within the candidate area. At each time point, the values ​​of the causal factors are extracted and matched with the sales values. Weights are set according to historical contributions to establish the correspondence between causal factors and sales changes. Based on the correspondence, causal residuals are calculated. Sales are predicted by causal factors and compared with actual sales to obtain residuals. The residuals are then superimposed with the abrupt change interval of sales slope in the time dimension to identify abnormal growth areas that cannot be explained by causal factors.

4. The AI-powered intelligent marketing method for identifying best-selling competitors based on product category and price, as described in claim 3, is characterized in that... Generating a cross-channel consistency spectrum based on causal residuals includes the following steps: The sales curves of each channel within the candidate region are extracted, and the local growth rate is calculated through a sliding window. The growth rate is converted into a phase signal, and the phase difference sequence is obtained with reference to the baseline curve. Based on the phase difference sequence, the synchronicity index between channels is calculated. The cross-channel consistency matrix is ​​obtained through correlation analysis and coherence function. The consistency matrix is ​​then matched point by point with the causal residuals in the time dimension to generate a consistency spectrum. The consistency spectrum is normalized and superimposed on the causal residual curve in the time dimension to form a joint anomaly signal. When the causal residual deviates positively and the consistency is low, the signal amplitude is enhanced to mark the abnormal growth area.

5. The AI-powered intelligent marketing method for identifying best-selling competitors based on product category and price, as described in claim 4, is characterized in that... When normalizing the consistency spectrum, the numerical range of the consistency spectrum is adjusted to match the numerical range of the causal residuals, and weighting coefficients are set during the superposition process, giving high weight to the causal residuals and low weight to the consistency spectrum.

6. The AI-powered intelligent marketing method for identifying best-selling competitors based on product category and price, as described in claim 4, is characterized in that... Constructing a dense tensor of user behavior under the constraint of anomalous signals includes the following steps: Under the constraint of abnormal signals, the user's order time is extracted from the transaction data of the candidate region, the time interval of consecutive orders is calculated, and the time interval density distribution is formed by kernel density estimation; Extract the geographic location information corresponding to the transaction data, map the order location to a two-dimensional space, and obtain the geographic density distribution through density clustering; By combining the time interval density distribution with the geographical density distribution, a user behavior density tensor is constructed, and the density peaks are compared with abnormal signals one by one in the tensor space.

7. The AI-powered intelligent marketing method for identifying best-selling competitors based on product category and price, as described in claim 1, is characterized in that... Generating a sales anomaly weight index under the joint constraints of user behavior density tensor and causal residuals includes the following steps: Feature vectors are extracted based on dense tensors of user behavior and causal residuals, a set of weight factors is constructed, and they are projected onto a unified numerical space through feature standardization. An abnormal sales weight index is generated based on the set of weighting factors. Abnormal signals are amplified by weighted superposition and exponential function. The output results are normalized to limit the range of values. The abnormal sales weight index is used to dynamically peel off the sales curves of candidate regions, weakening the abnormal growth part with a decay coefficient and retaining the normal trend part. The abnormal sales weight index is written back to the competitor profile to correct the original best-selling weight and avoid competitor identification bias caused by false sales inflation.

8. The AI-powered intelligent marketing method for identifying best-selling competitors based on product category and price, as described in claim 7, is characterized in that... The process of converting the sales curve into a time-spectrum signal, driven by an abnormal sales weighting index, includes the following steps: Driven by the abnormal sales weight index, the sales curves of candidate regions are weighted and adjusted, and divided into adjacent time windows to form the input signal for spectrum analysis. Applying short-time Fourier transform to the time window yields the spectral distribution of low-frequency trends, mid-frequency fluctuations, and high-frequency anomalies. Low-frequency trends and mid-frequency fluctuations are directly incorporated into the sales weighting, and the sales anomaly weighting index is used to attenuate the amplitude of abnormal signals within the high-frequency anomaly range. The low-frequency trend and mid-frequency fluctuation are merged and updated to the best-selling weights in the competitor profile, and the attenuated high-frequency anomalies are used as monitoring indicators and written back to the competitor profile.

Images