Electronic commerce promotion method and system based on cloud computing

Abstract

The invention relates to the technical field of e-commerce, in particular to an e-commerce promotion method and system based on cloud computing, and the method comprises the following steps: building a user behavior sequence through log collection, extracting a sudden change point through dynamic time warping, generating a dynamic parameter through segmented integral, calculating a difference degree through gray correlation analysis, and reconstructing a weight set. And flexibly scheduling advertisement resource distribution, and generating a delivery instruction. According to the method, user click stream data is collected, a behavior sequence model is constructed, continuous dynamic characteristics of user browsing tracks are captured, time sequence relevance is analyzed, fluctuation peak points of user behavior paths are aligned, interest mutation nodes are captured, abnormal behavior modes are recognized, and implicit preferences and explicit feedback parameters in the interaction process are quantified. Establishing an associated feature analysis mechanism, resetting advertisement weight parameters, matching real-time requirements of users with putting strategies, mapping weights to distributed advertisement nodes, dynamically scheduling computing resources, adapting display requirements and resource allocation, and forming a closed-loop optimization link.

Description

Technical Field

[0001] The present invention relates to the technical field of e-commerce, and particularly to an e-commerce promotion method and system based on cloud computing. Background Art

[0002] The technical field of e-commerce includes information management and service means for the whole process of displaying, selling, paying, and logistics of goods and services through a network platform. The core content of this technical field involves the construction and operation and maintenance of an electronic trading platform, the docking and management of an online payment system, the collection and analysis of consumer data, the automatic coordination of the supply chain, and the implementation of network marketing strategies. E-commerce involves various business models such as B2B, B2C, and C2C at the application level, and its overall development depends on the information network infrastructure, the security guarantee mechanism of the trading system, and the continuous optimization of the front-end interaction experience. On this basis, e-commerce is deeply integrated with emerging information technologies such as cloud computing, big data, and artificial intelligence, forming a cross-regional, multi-terminal, and scalable business operation system.

[0003] Among them, the e-commerce promotion method based on cloud computing refers to a technical solution that uses distributed computing resources and a network platform to achieve the wide dissemination and targeted marketing of e-commerce content. This patent theme mainly focuses on the marketing investment link in e-commerce activities, covering technical matters such as commodity information collection, user behavior analysis, advertising placement strategy generation, and multi-channel content distribution. Specifically, based on a cloud computing platform, by collecting and aggregating user access behavior logs, establishing user portraits in combination with commodity attribute tags, and then executing the matching and distribution process of advertising content and target users according to classification logic, while supporting cross-platform synchronous display and status feedback recovery. The above process is supported by the elastic resource scheduling and high-concurrency processing capabilities of the cloud platform to complete the data-driven and intelligent matching of the entire e-commerce promotion process.

[0004] The prior art relies on static user portraits and preset rules to perform advertisement matching, lacking the ability to continuously track real-time behavior sequences, resulting in lagging strategy updates in scenarios of interest drift. For example, the tag system of user behavior logs is generated based on fixed time windows, making it difficult to capture sudden changes in interests within short cycles, causing a mismatch between advertisement placement and current demands. Most existing advertisement weight allocation mechanisms adopt a periodic batch update mode, unable to respond promptly to high-frequency behavior fluctuations, and prone to generating ineffective exposures when user preferences switch rapidly. Traditional resource scheduling strategies perform linear expansion based on preset capacities, lacking sufficient dynamic adaptability between resource allocation and task queues in the face of sudden traffic peaks and valleys, which may lead to overload or idleness of local nodes. In addition, the split analysis of positive and negative behavior data easily ignores the dynamic coupling relationship between the two during the interaction process. For example, there may be a temporal correlation between page skip behavior and dwell time, and it is difficult for the prior art to mine such composite features, resulting in deviations in user intention judgment. The above problems will reduce the accuracy of advertisement reach and increase operating costs. Summary of the Invention

[0005] The object of the present invention is to solve the drawbacks existing in the prior art, and to propose an e-commerce promotion method and system based on cloud computing.

[0006] To achieve the above object, the present invention adopts the following technical solutions: An e-commerce promotion method based on cloud computing, comprising the following steps: S2: Input the user behavior sequence into the dynamic time warping algorithm, perform sliding window alignment on the Euclidean distance and time interval difference of adjacent jump paths, extract the fluctuation peak points that exceed the threshold three times continuously, and generate a mutation point sequence; S3: Invoke the positive behavior data and negative behavior data of the commodity interaction log, and perform piecewise integral operations on the number of collections, dwell time on the detail page, page skip frequency, and fast sliding frequency based on the mutation point sequence to generate positive dynamic parameters and negative dynamic parameters; S4: Input the positive dynamic parameters and the negative dynamic parameters into the grey relational analysis method, calculate the difference in the degree of association between the two within three consecutive time windows, and when the difference exceeds the preset threshold, perform a priority reset operation on the advertisement content weight set to generate a reconstructed weight set; S5: Based on the reconstructed weight set, invoke the elastic resource scheduling algorithm of the cloud server, allocate the computing resources and storage resources of the advertisement placement node, and map them to the advertisement display queue to generate an advertisement resource distribution instruction.

[0007] As a further solution of the present invention, the mutation point sequence includes a fluctuation peak point timestamp, a fluctuation amplitude value, and an adjacent peak point interval. The positive dynamic parameters are specifically the integral value of the collection times and the integral value of the stay duration on the detail page. The negative dynamic parameters are specifically the integral value of the page skip frequency and the integral value of the fast sliding times. The reconstructed weight set includes a reset weight value, a priority label, and an advertisement identifier. The advertisement resource distribution instruction specifically refers to a computing node allocation strategy, a storage node allocation strategy, and an advertisement display order list.

[0008] As a further solution of the present invention, the steps for obtaining the mutation point sequence are specifically as follows: S201: Obtain adjacent jump paths in the user behavior sequence, calculate the spatial difference value of adjacent path coordinate points, extract the timestamp difference of the corresponding path, combine the Euclidean distance and the time interval difference in the path order to form a difference pair, and generate a path difference pair sequence; S202: Based on the path difference pair sequence, set a fixed length of the sliding window, perform an alignment operation on the sum of the dispersion degree of the Euclidean distance difference and the absolute value of the time interval difference within the window, calculate the comprehensive metric value of the fluctuation intensity within the window, and use the formula: ; Calculate the window fluctuation intensity through the operation, compare the metric value with a preset alignment threshold, and screen the windows exceeding the threshold and mark them as candidate fluctuation peak points; Among them, represents the fluctuation intensity of the th window, represents the variance value of the Euclidean distance between adjacent paths within the window, represents the sum of the absolute values of the time interval differences between adjacent paths within the window, is the cumulative value of the number of fluctuations within the current window, is the fixed length of the sliding window; S203: Call the candidate fluctuation peak points, count the continuous occurrence times of each peak point corresponding to the window, set a continuous times threshold, and eliminate the peak points that do not meet the continuous super-threshold condition to generate a mutation point sequence.

[0009] As a further solution of the present invention, the steps for obtaining the positive dynamic parameters and the negative dynamic parameters are specifically as follows: S301: Call the positive behavior data and negative behavior data of the commodity interaction log. Based on the timestamp of the mutation point sequence, divide the collection times into intervals according to adjacent mutation points, truncate the stay duration on the detail page into discrete segments according to the mutation points, and classify the page skip frequency and the sliding times according to time windows to generate a segmented data set; S302: Perform piecewise integral operations on the number of collections, the dwell time on the detail page, the page skip frequency, and the number of swipes for each segment in the segmented dataset, using the formula: ; Perform the operation to obtain the integral value, calculate the difference between the integral results of the positive behavior data and the negative behavior data, and generate an integral vector; Among them, represents the dynamic integral value of the type of behavior, represents the number of collections in the segment, represents the dwell time on the detail page in the segment, represents the time interval of the mutation point in the segment, represents the page skip frequency in the segment, represents the number of swipes in the segment, represents the user activity coefficient in the segment, represents the balance factor of the interaction density and the behavior type, represents the total number of segments, represents the start timestamp of the segment; S303: Separate the positive integral and the negative integral according to the positive and negative signs of the integral vector, normalize the positive integral using the maximum-minimum method, normalize the negative integral using the standard deviation method, and allocate the time weight factor based on the segment duration ratio to generate the positive dynamic parameter and the negative dynamic parameter.

[0010] As a further solution of the present invention, the steps for obtaining the reconstructed weight set are specifically as follows: S401: Call the positive dynamic parameter and the negative dynamic parameter, based on the parameter sequences of three consecutive time windows, align the positive and negative parameters of each window point by point, calculate the correlation coefficient, and at the same time take the mean value within the window as the correlation degree, integrate the data of the three windows, and generate a window correlation degree set; S402: Extract the adjacent window correlation degrees from the window correlation degree set, calculate the absolute difference between the subsequent window and the previous window, obtain the differences between the first and second windows and between the second and third windows, compare the two and take the maximum value as the current difference degree, and at the same time compare the current difference degree with the preset difference degree threshold. If it exceeds the threshold, generate a difference degree status flag; S403: Extract the original priority of the advertisement content weight set according to the difference degree status flag, reassign the weight values according to preset rules, arrange them in descending order, overwrite the original set, and generate a reconstructed weight set.

[0011] As a further solution of the present invention, the step of obtaining the advertisement resource distribution instruction is specifically as follows: S501: Invoke the reconstructed weight set, extract the computing resource requirement value and storage resource requirement value of the advertisement placement node, combine the node load value and delay coefficient in the elastic resource scheduling algorithm of the cloud server, and use the formula: ; Calculate to obtain the node resource allocation parameter value, and generate a resource allocation parameter set; Wherein, represents the resource allocation parameter of node , is the dynamic reconstruction weight of node in the reconstructed weight set, is the computing resource requirement value of node , is the storage resource requirement value of node , is the current load value of node , is the delay coefficient of node , represents the reconstruction weight correlation item, represents the delay correlation item; S502: Based on the resource allocation parameter set, compare the node resource allocation parameter value with the priority threshold of the advertisement display queue numerically, eliminate the parameter values lower than the priority threshold, and intercept the parameter value sequence within the queue length limit according to the time window constraint to generate a node resource adjustment queue; S503: Invoke the node resource adjustment queue, bind the computing resource requirement value and storage resource requirement value corresponding to each parameter value in the queue item by item according to the time sequence number of the advertisement display queue, and generate an advertisement resource distribution instruction.

[0012] As a further solution of the present invention, the method further includes: S1: Obtain user click stream data through the cloud log collection node, perform cleaning and alignment operations on the page jump path, stay duration, and product category label, and construct a user behavior sequence based on the hidden Markov model; The user behavior sequence is specifically the jump timestamp, theme encoding, and page level.

[0013] As a further solution of the present invention, the step of obtaining the user behavior sequence is specifically as follows: S101: Collect the original user clickstream data in the cloud logs, extract the page jump path, stay duration, and product category label fields, filter the missing path nodes and abnormal stay duration values, match the product category labels with the page paths, and generate standardized clickstream data; S102: Based on the standardized clickstream data, align the timestamps of the page jump paths and product category labels, unify the time precision of different terminals, accumulate the stay duration segment by segment according to the path nodes, extract the continuous path, duration, and label combinations under the same user ID, and generate a set of behavior trajectory parameters; S103: Invoke the path jump frequency, average stay duration, and label correlation degree in the set of behavior trajectory parameters, calculate the state transition probability matrix and the stay duration distribution density, map the path nodes to the state variables of the hidden Markov model, establish the corresponding relationship between the state transition and the observation sequence, and generate a user behavior sequence model.

[0014] An e-commerce promotion system based on cloud computing, the e-commerce promotion system based on cloud computing is used to execute the above-mentioned e-commerce promotion method based on cloud computing, and the system includes: A behavior sequence construction module, which is used to obtain user clickstream data through a cloud log collection node, perform data cleaning and timestamp alignment operations on the page jump path, stay duration, and product category labels, input the cleaned data into a hidden Markov model to generate a user behavior sequence, and transfer the user behavior sequence to a mutation point extraction module; A mutation point extraction module, which is used to call the user behavior sequence, perform a sliding window alignment operation on the Euclidean distance and time interval difference between adjacent jump paths based on the dynamic time warping algorithm, extract the fluctuation peak points that exceed the preset threshold three times in a row, generate a mutation point sequence, and transfer the mutation point sequence to a dynamic parameter generation module; A dynamic parameter generation module, which is used to obtain the positive behavior data and negative behavior data in the product interaction logs, perform piecewise integral operations on the number of collections, stay duration on the detail page, page skip frequency, and fast sliding times based on the mutation point sequence, generate positive dynamic parameters and negative dynamic parameters, and transfer the positive dynamic parameters and the negative dynamic parameters to a weight reconstruction module; A weight reconstruction module, which is used to call the positive dynamic parameters and the negative dynamic parameters, calculate the difference value of the correlation degree between the two in three consecutive time windows based on the grey relational analysis method, perform a priority reset operation on the advertising content weight set when the difference value exceeds the dynamic threshold, generate a reconstructed weight set, and transfer the reconstructed weight set to a resource scheduling mapping module; A resource scheduling mapping module, which is used to call the elastic resource scheduling algorithm of the cloud server based on the set of reconstructed weights, dynamically adjust the allocation ratio of computing resources and storage resources of the advertisement placement nodes, map the adjusted resource allocation scheme to the advertisement display queue, generate an advertisement resource distribution instruction and output it to the cloud execution node.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In the present invention, by collecting user clickstream data and constructing a hidden Markov behavior sequence, continuous dynamic modeling of the user browsing trajectory is realized, and the temporal correlation of behavior prediction is improved. The dynamic time warping algorithm is used to perform window alignment on the path fluctuation peak points to capture the critical nodes of user interest mutation and enhance the sensitivity of abnormal behavior recognition. By combining the piecewise integral operations of positive and negative behavior parameters, the correlation characteristics of implicit preferences and explicit feedback in the interaction process are quantified, and the misjudgment risk caused by single-dimensional indicators is reduced. The dynamic reset of advertisement weights is carried out using the difference in grey correlation degree, and a matching mechanism between the placement strategy and the real-time needs of users is established to avoid the response lag caused by fixed weight allocation. Based on the elastic resource scheduling algorithm, the reconstructed weights are mapped to distributed advertisement nodes to achieve precise adaptation of computing resources and display requirements, and alleviate the load imbalance problem in high-concurrency scenarios. The above steps form a closed-loop optimization link, enabling the advertisement content to be dynamically adjusted following the fluctuations of user behavior, improving the placement efficiency and conversion rate while reducing redundant resource consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the working process of the present invention; Figure 2 It is a flowchart of the steps for obtaining the user behavior sequence of the present invention; Figure 3 It is a flowchart of the steps for obtaining the mutation point sequence of the present invention; Figure 4 It is a flowchart of the steps for obtaining the positive dynamic parameter and negative dynamic parameter of the present invention; Figure 5 It is a flowchart of the steps for obtaining the set of reconstructed weights of the present invention; Figure 6 It is a flowchart of the steps for obtaining the advertisement resource distribution instruction of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0018] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0019] Embodiment 1 Please refer to Figure 1 , the present invention provides a technical solution: an e-commerce promotion method based on cloud computing, including the following steps: S1: Obtain user clickstream data through cloud log collection nodes, perform cleaning and alignment operations on the page jump path, stay duration, and product category tags, and construct a user behavior sequence based on the hidden Markov model; S2: Input the user behavior sequence into the dynamic time warping algorithm, perform sliding window alignment on the Euclidean distance and time interval difference of adjacent jump paths, extract the fluctuation peak points that exceed the threshold three times in a row, and generate a mutation point sequence; S3: Invoke the positive behavior data and negative behavior data of the product interaction log, perform piecewise integral operations on the number of favorites, the stay duration on the detail page, the page skip frequency, and the fast sliding frequency based on the mutation point sequence, and generate positive dynamic parameters and negative dynamic parameters; S4: Input the positive dynamic parameters and negative dynamic parameters into the grey relational analysis method, calculate the difference in the degree of association between the two in three consecutive time windows, and when the difference exceeds the preset threshold, perform a priority reset operation on the advertising content weight set to generate a reconstructed weight set; S5: Based on the reconstructed weight set, invoke the cloud server elastic resource scheduling algorithm, allocate the computing resources and storage resources of the advertising placement node, and map them to the advertising display queue to generate an advertising resource distribution instruction.

[0020] The user behavior sequence specifically includes jump timestamps, theme codes, and page hierarchy values. The mutation point sequence includes fluctuation peak point timestamps, fluctuation amplitude values, and the interval between adjacent peak points. The positive dynamic parameters specifically include the integral value of the number of favorites and the integral value of the stay duration on the detail page. The negative dynamic parameters specifically include the integral value of the page skip frequency and the integral value of the fast sliding frequency. The reconstructed weight set includes reset weight values, priority tags, and advertising identifiers. The advertising resource distribution instruction specifically refers to the computing node allocation strategy, the storage node allocation strategy, and the advertising display order list.

[0021] Please refer to Figure 2, the specific steps for obtaining the user behavior sequence are as follows: S101: Collect the original user clickstream data in the cloud log, extract the page jump path, stay duration, and product category label fields, filter out the missing path nodes and abnormal stay duration values, match the product category label with the page path, and generate standardized clickstream data; Collect the original user clickstream data from cloud logs. The specific action is to retrieve log files within a specified time range from the server log system storing user activity records, such as the object storage service (OSS) or database (e.g., ClickHouse) deployed on Alibaba Cloud or Tencent Cloud. For example, access logs for the past 24 hours are retrieved. These logs contain original fields such as timestamps (e.g., '2025-04-10 10:30:15.123'), user identifiers (e.g., 'User_A1B2'), accessed page URLs (e.g., ' / product / detail / item123'), request types (e.g., 'GET'), user agents (e.g., browser information), IP addresses, etc. Then, extract key information from these original fields and identify the URL fields representing the page jump path. For example, extract an access sequence like ' / home' -> ' / category / electronics' -> ' / product / list / tv' -> ' / product / detail / tv_brand_X_model_Y'. At the same time, extract the start and end timestamps corresponding to each page access and calculate the page dwell time. For example, if the start timestamp of a user on the ' / product / list / tv' page is '10:31:05.200' and the timestamp of jumping to ' / product / detail / tv_brand_X_model_Y' is '10:31:45.700', then the dwell time is calculated as 40.5 seconds. Further, extract the possible product category label fields in the logs. This field may be directly recorded in the log (e.g., 'category=TV'), or it may need to be parsed and mapped through the accessed URL path. For example, through a preset URL-category mapping rule, map ' / product / detail / tv_brand_X_model_Y' to the 'TV' category label. Subsequently, clean the extracted data by filtering out records with incomplete path node information, such as URLs missing key page identifiers, and abnormal dwell time values. Set a reasonable dwell time range, for example, filter out records below 1 second or above 1800 seconds (30 minutes), considering these as invalid clicks or hanging behaviors. For example, if a record shows a dwell time of 3000 seconds on a product detail page, then filter it out. Finally, match and associate the product category labels corresponding to valid path nodes based on timestamps and user identifiers. For example, confirm '10:31:45.The product category for user 'User_A1B2' who accesses ' / product / detail / tv_brand_X_model_Y' at '700' is 'TV'. The processed page jump path, dwell time, and product category label are combined into a structured record to form standardized clickstream data. For example, a record set with the format {UserID: 'User_A1B2', Timestamp: '10:31:45.700', PagePath: ' / product / detail / tv_brand_X_model_Y', DwellTime: 40.5, Category: 'TV'} is generated.

[0022] S102: Based on the standardized clickstream data, the timestamps of the page jump paths and the product category labels are aligned to unify the time accuracy of the differentiated terminals. The dwell time is accumulated by segment by path node, and the continuous path, duration, and label combination under the same user ID is extracted to generate a set of behavior trajectory parameters. First, we address the differences in log timestamp accuracy that may exist between different sources or terminal devices. For example, we align the millisecond timestamps recorded by the mobile app (such as '10:31:45.715') and the second timestamps recorded by the web (such as '10:31:46') to second-level accuracy, and use a unified rounding or rounding rule. For example, we round down to the nearest second, so that '10:31:45.715' becomes '10:31:45' and '10:31:46' remains unchanged, ensuring that the time base for subsequent processing is consistent. Next, we align the consecutive access records of the same user on the same page path node and accumulate the duration of these records. For example, if user 'User_A1B2' has two consecutive records on the ' / product / list / tv' page with durations of 15.2 seconds and 25.3 seconds respectively, we merge these two records and record the total duration as In seconds, the associated timestamp adopts the timestamp of the last record, and the product category labels are merged or their consistency is confirmed accordingly. Subsequently, all the processed records of this user are concatenated in chronological order according to the user ID (such as 'User_A1B2'), and a continuous page jump path sequence, a merged dwell time sequence, and a corresponding product category label sequence are extracted. For example, the behavior trajectory of user 'User_A1B2' is formed: [(' / home', 5.0, 'None'), (' / category / electronics', 10.1, 'Electronics'), (' / product / list / tv', 40.5, 'TV'), (' / product / detail / tv_brand_X_model_Y', 65.8, 'TV'), (' / cart', 30.2, 'None')], where each tuple represents (path node, cumulative dwell time, product category label). The trajectory parameters of all users are combined to generate a set of behavior trajectory parameters.

[0023] S103: Invoke the path jump frequency, average dwell time, and label correlation degree in the set of behavior trajectory parameters, calculate the state transition probability matrix and the dwell time distribution density, map the path nodes to the state variables of the hidden Markov model, establish the corresponding relationship between state transitions and observation sequences, and generate a user behavior sequence model.

[0024] First, calculate the state transition probability. Traverse the adjacent path jump instances of all users in the set of behavior trajectory parameters, and count the frequency of jumping from path node to path node , and the total frequency of jumping from path node to all other nodes . Then the state transition probability is calculated as . For example, it is statistically found that the frequency of jumping from ' / product / list / tv' to ' / product / detail / tv_brand_X_model_Y' is 500 times, and the total frequency of jumping from ' / product / list / tv' to all other pages is 1000 times. Then the transition probability is , calculate the transition probabilities between all path nodes to form a state transition probability matrix. Meanwhile, calculate the residence duration distribution density of each path node (state), and call the residence duration data of each path node in the behavioral trajectory parameter set. For example, for the path ' / product / list / tv', the collected residence duration data is {40.5, 35.2, 55.1,...}. Use methods such as Kernel Density Estimation (KDE) to fit the probability density function of these duration data , this function describes the probability density of the residence duration in state and analyze the correlation between product category labels and path nodes, calculate the frequency or conditional probability of a specific label appearing at the path node , for example, calculate the probability that the 'TV' label appears under the path ' / product / list / tv'. After that, map each unique page path node (such as ' / home', ' / product / list / tv') to a hidden state in the HMM, and use the calculated residence duration distribution density and label correlation as part or the basis of the observation probability (emission probability) of the state . The observation sequence can be the pair of (residence duration, product category label) actually observed, such as (40.5s, 'TV'). Establish the correspondence between the state transition probability matrix and the observation sequence (residence duration distribution, label correlation), and clarify that the probability of state transferring to state is determined by , and the probability of state generating an observation value (such as residence duration , label ) is jointly determined by and etc. Finally, complete the construction of the user behavior sequence model and .

[0025] Please refer to Figure 3 , the specific steps for obtaining the mutation point sequence are as follows: S201: Obtain adjacent jump paths in the user behavior sequence, calculate the spatial difference value of adjacent path coordinate points, extract the time difference of the corresponding paths, combine the Euclidean distance and time interval difference in path order to form a difference pair, and generate a path difference pair sequence Using the path sequence in the set of behavioral trajectory parameters, extract two page jump paths that are adjacent in time. For example, if there is ' / product / list / tv' -> ' / product / detail / tv_brand_X_model_Y' -> ' / cart' in the sequence, then the adjacent jump path pairs are (' / product / list / tv', ' / product / detail / tv_brand_X_model_Y') and (' / product / detail / tv_brand_X_model_Y', ' / cart'). To calculate the spatial difference value of path coordinate points, each page path needs to be mapped to a multi-dimensional space coordinate in advance. This mapping can be achieved through vectorization of page content (such as obtaining vectors by processing page text elements using TF-IDF or BERT models) or graph embedding algorithms based on page link relationships (such as Node2Vec). Assume that the mapped coordinate of ' / product / list / tv' is , and the mapped coordinate of ' / product / detail / tv_brand_X_model_Y' is , then the spatial difference value between them, that is, the Euclidean distance, is calculated as ; At the same time, extract the time difference between the timestamps corresponding to these two adjacent paths, that is, the access time of the latter path node minus the access time of the former path node. This represents the residence duration or jump interval at the first path node. For example, if the access timestamp (the time of entering the page) of ' / product / list / tv' is '10:31:05' and the access timestamp of ' / product / detail / tv_brand_X_model_Y' is '10:31:45', then the time difference seconds. Combine the calculated Euclidean distance and the time interval difference seconds into a difference pair in the order of the original path jumps . Repeat this process for all adjacent paths in the user behavior sequence to generate a sequence of path difference pairs. For example , where is the Euclidean distance of the path for the th jump, and is the time interval for the th jump.

[0026] S202: Based on the sequence of path difference pairs, set the fixed length of the sliding window, and perform an alignment operation on the sum of the dispersion degree of the Euclidean distance differences and the absolute values of the time interval differences within the window, and calculate the comprehensive measure value of the fluctuation intensity within the window. Use the formula: ; Obtain the window fluctuation intensity through calculation, compare the metric value with a preset alignment threshold, and screen the windows exceeding the threshold and mark them as candidate fluctuation peak points; Among them, represents the fluctuation intensity of the -th window, represents the variance value of the Euclidean distance between adjacent paths within the window, represents the sum of the absolute values of the differences in the time intervals between adjacent paths within the window, is the cumulative value of the number of fluctuations within the current window, is the fixed length of the sliding window; Set a sliding window with a fixed length, and denote the window length as , for example, set , which means analyzing 10 consecutive path differences each time. The window slides through the entire sequence in turn. For the -th window, this window contains the -th to the -th difference pairs, that is, , calculate the dispersion degree of the Euclidean distance difference within the window. Specifically, calculate the variance of these Euclidean distance values. The calculation formula is , where is the average value of the Euclidean distances within the window. For example, the values within the window are {0.7, 0.8, 0.6, 0.9, 0.7, 0.5, 0.8, 0.7, 0.9, 0.6}, calculate its average value , and the variance ; At the same time, calculate the sum of the absolute values of the time interval differences within the window, and denote it as . Note that in the formula description refers to the sum of the absolute values, rather than "performing an alignment operation on the absolute value of the time interval difference" in the original text. We will execute according to the formula description. Assume that the values within the window are {40, 35, 50, 30, 45, 60, 25, 38, 32, 48} (unit: seconds), then seconds, obtain the cumulative value of the number of fluctuations within the current window, which is the length of the window, unless there is a special definition (such as only counting the times). Here, we set , and then, call the fluctuation intensity calculation formula: Here, is added with an absolute value symbol in the formula, but It is the sum of absolute values and is always non-negative, so the absolute value sign can be omitted. The formula parameters are explained in detail: Representative The fluctuation intensity of a window. A larger value indicates a more drastic change in user behavior within the window. Represents the variance of the Euclidean distance between adjacent paths within the window, which measures the stability of the path space change. The larger the value, the more unstable it is. It represents the sum of the absolute values of the time interval differences between adjacent paths within the window, and measures the total time span or activity within the window; The cumulative value of the number of fluctuations in the current window (here equal to the window length ), through its square root Scale the intensity of fluctuations; The sliding window is fixed in length and is normalized as the denominator. The operation logic is to convert the instability of spatial changes (variance ) and time span ( ) and by the number of data points in the window ( ) is enlarged and then the window length ( ) is normalized to obtain an indicator that comprehensively measures the severity of the behavior fluctuation within the window, and the example values are used for calculation: ; The formula is beneficial in that it combines the spatial variation of the path coordinates ( ) and the total amount of time intervals ( ), and considering the number of data points within the window ( ) and the window size ( ), which can more comprehensively capture the comprehensive fluctuations of user behavior within a short time window and identify potential time points when behavior patterns change significantly. Next, set a fluctuation intensity alignment threshold The setting of the threshold can refer to the historical data The distribution of values, such as taking historical The 90th percentile of the value, or set an empirical value based on business needs. Suppose that by analyzing historical data, set , the calculated window fluctuation intensity With threshold For comparison, , so the If a window exceeds the threshold, the window is marked as a candidate fluctuation peak point, and the calculation and comparison process is repeated for all windows to filter out all windows that exceed the threshold and mark them as a candidate fluctuation peak point set.

[0027] This result shows that the The user behavior fluctuation intensity within a window is relatively high, reaching a level that requires attention, which provides a candidate basis for subsequent identification of mutation points. The calculated value is a candidate signal.

[0028] S203: Call the candidate fluctuation peak points, count the consecutive occurrence times of each peak point corresponding window, set the consecutive occurrence times threshold, eliminate the peak points that do not meet the condition of consecutive exceeding the threshold, and generate a mutation point sequence.

[0029] For example, the window numbers marked as candidate peak points are {5, 6, 7, 12, 18, 19, 20, 21, 25,...}. Check whether these candidate peak points appear consecutively in the sequence, and count the consecutive occurrence times of each peak point corresponding window. For example, windows 5, 6, and 7 appear consecutively, with a consecutive occurrence times of 3; window 12 appears alone, with a consecutive occurrence times of 1; windows 18, 19, 20, and 21 appear consecutively, with a consecutive occurrence times of 4; window 25 appears alone, with a consecutive occurrence times of 1. Set a consecutive occurrence times threshold , which is used to filter out accidental and short-duration fluctuations, ensuring that the identified mutation points have a certain stability. The setting of the threshold can be based on experience. For example, it is required that at least 3 windows appear consecutively to be considered a valid mutation. Let , compare the consecutive occurrence times of each candidate fluctuation peak point with . For windows 5, 6, and 7, the consecutive occurrence times of 3 meet , and these peak points are retained; for window 12, the consecutive occurrence times of 1 do not meet , and this peak point is eliminated; for windows 18, 19, 20, and 21, the consecutive occurrence times of 4 meet , and these peak points are retained; for window 25, the consecutive occurrence times of 1 do not meet , and this peak point is eliminated. Each window (or the time point it represents, such as the central time point or the end time point of the window) in the retained consecutive peak point sequence (such as windows 5 - 7, windows 18 - 21) is confirmed as a behavior mutation point. Collect the timestamps or serial numbers of all confirmed mutation points to generate the final mutation point sequence, such as {Timestamp_peak5, Timestamp_peak6, Timestamp_peak7, Timestamp_peak18, Timestamp_peak19, Timestamp_peak20, Timestamp_peak21,...}.

[0030] Please refer to Figure 4 , the specific steps for obtaining the positive dynamic parameter and the negative dynamic parameter are as follows: S301: Invoke the positive and negative behavior data of the product interaction logs. Based on the timestamps of the mutation point sequences, divide the number of favorites into intervals according to adjacent mutation points, truncate the dwell time on the detail page into discrete segments at the mutation points, and classify the page skip frequency and the number of swipes according to time windows to generate a segmented data set; Invoke the product interaction logs, which record the specific interaction behaviors of users with products. It is necessary to distinguish the positive and negative behavior data among them. Positive behaviors refer to behaviors where users show interest or intention, such as product favorites (Favorite), adding to the shopping cart (AddToCart), purchasing (Purchase), and long-term browsing of the detail page (LongDwellTimeonDetailPage); negative behaviors refer to behaviors where users show disinterest or avoidance, such as page skipping (Skip), fast swiping (FastScroll), and quickly returning from the detail page (QuickBackfromDetailPage). Based on the timestamps included in the mutation point sequence {Timestamp_mut1, Timestamp_mut2,...}, segment these interaction logs by time. The time range of the first segment is from the start of the sequence to the first mutation point , the second segment is from to , and so on. The time range of the th segment is (assuming is the start time of the sequence). For each segment , count the various interaction behavior data within this time interval: calculate the number of favorites , for example, within the interval, the user favorited the product 2 times; calculate the dwell time on the detail page , accumulate the dwell times of all visits to the detail page within this interval, or truncate the dwell time across mutation points according to the mutation point timestamps, only including the part of the time that falls within this interval. For example, if a certain detail page visit starts from and ends at , then the duration belonging to segment is 10 seconds; count the page skip frequency , for example, the user skipped the recommended products 5 times within this interval; count the number of swipes , for example, the user performed 30 swipe operations within this interval. Organize the { } data and the timestamp information of the segments statistically calculated for all segments into a segmented data set.

[0031] Table 1 Example Table of Segmented Interaction Data

[0032] As shown in Table 1, the table lists the statistical data of user interaction behaviors within three consecutive time segments divided according to mutation points.

[0033] S302: Perform piecewise integral operations on the number of collections, the duration of staying on the detail page, the page skip frequency, and the number of slides for each segment in the segmented dataset, using the formula: ; Obtain the integral value through the operation, calculate the difference between the integral results of the positive behavior data and the negative behavior data, and generate an integral vector; Among them, represents the dynamic integral value of the th type of behavior, represents the number of collections in the th segment, represents the total duration of staying on the detail page in the th segment, represents the time interval of the mutation point in the th segment, represents the page skip frequency in the th segment, represents the number of slides in the th segment, represents the user activity coefficient in the th segment, represents the balance factor of interaction density and behavior type, represents the total number of segments, represents the th segment's start timestamp; Perform piecewise integral operations on the interaction data for each segment in the segmented dataset to obtain the dynamic integral value of each behavior type (positive / negative), and the formula used is: According to the previous analysis, a more reasonable interpretation of this formula is: ; Detailed description of the formula parameters: represents the cumulative dynamic integral value of the th type of behavior (for example, s can be "overall" or distinguish "positive", "negative"); is the index of the segment, ranging from 1 to the total number of segments ; is the number of collections within the th segment; is the total duration of staying on the detail page within the th segment (in seconds); is the time length (in seconds) of the th segment, that is ; is the The page skip frequency within a segment; is the number of swipes within a segment; is the user activity coefficient of a segment, used to standardize the impact of negative behaviors, and its calculation method can be set as the ratio of the total number of interactions within the segment ( ) to the segment duration , and then normalized, for example, normalized to the interval [0.1, 1] to avoid the denominator being zero or too small. Suppose the of segment 1 is normalized to 0.8, the of segment 2 is 0.5, and the of segment 3 is 0.9; is the balance factor between interaction density and behavior type, used to adjust the weight of negative behaviors (skipping, swiping) relative to positive behaviors (collecting, staying), and its setting can be determined according to the experimental results. For example, set , aiming to balance the impact of the two behaviors on the final score; the summation symbol means adding up the values within the parentheses calculated for all segments to obtain the final total score . The first term of the formula quantifies the contribution of positive behaviors by multiplying the product of the number of collections and the stay duration (representing positive engagement) by the square root of the segment duration (giving longer segments higher weight, but the effect is weakened by taking the square root), and taking the absolute value to ensure non - negativity; the second term of the formula quantifies the net negative behavior by subtracting the number of swipes from the skip frequency ( may be negative, indicating that swiping dominates), dividing by the activity coefficient for standardization, and multiplying by the balance factor to adjust its impact weight; the two are added and then summed across segments to obtain an integral value that comprehensively reflects the dynamic behavior tendency of the user during the entire observation period.

[0034] The benefit of the formula is that it combines various positive and negative user interaction behaviors, and through segment - by - segment calculation and time weighting ( ), as well as activity standardization and balance factor adjustment, it can dynamically and quantitatively evaluate the comprehensive participation and interest changes of users in each stage divided by behavior mutation points. Using the data in Table 1 for an example calculation ( , ): First, calculate each segment and the assumed : Segment 1: , ; Segment 2: , ; Segment 3: , ; Calculate the contribution value of each segment: Contribution of Segment 1 ; Contribution of Segment 2 ; Contribution of Segment 3 ; Total score ; The operation obtains an integral value of , this integral value combines the positive and negative behavior performances of the user in the three segments. Next, it is necessary to distinguish whether the calculation is based on the positive behavior log or the negative behavior log, or as in this example, combine the positive and negative indicators to calculate the total score, and then perform separation or difference calculation. Assuming that a comprehensive integral is calculated here, further processing is required to obtain the positive and negative parameters. If the pure positive indicators (such as only ) and pure negative indicators (such as only ) are applied to calculate and using a similar logic, and then perform a difference calculation to generate an integral vector, for example, obtaining the vector . The here is an example of a comprehensive score. This result indicates that the user's overall interaction shows a strong positive trend during this period (because the integral is much greater than 0). This integral value will be used as the basis for the next normalization and weighting.

[0035] S303: Separate the positive integral and the negative integral according to the positive and negative signs of the integral vector, normalize the positive integral using the maximum-minimum method, normalize the negative integral using the standard deviation method, and allocate the time weight factor based on the proportion of the segment duration to generate the positive dynamic parameter and the negative dynamic parameter.

[0036] Separate the positive integral and the negative integral according to the positive and negative signs (or sources) of the integral value. For example, we assume that through separation calculation, we obtain and (the sum of these two values is approximately equal to the total integral calculated previously). For the positive integral , perform normalization processing using the maximum-minimum (Min-Max) method. It is necessary to determine the maximum value and the minimum value of the positive integral within a batch of users or a time period. Assuming and , then the normalized positive integral ; Negative integral Using the standard deviation (Z-score) method for normalization, it is necessary to calculate the mean of the negative integral for a batch of users or within a certain time period and the standard deviation , assuming and , then the normalized negative integral ; Next, based on the proportion of each segment duration to the total duration, allocate the time weight factor and calculate the total duration seconds; Calculate the proportion of the duration of each segment , for example , , ; This time weight factor can be used to weightedly adjust the final normalization parameter, or as implied in the original text, it may be used to generate a total time weight factor, for example, adjusted according to the proportion of the duration of the segment where positive behavior mainly occurs or the segment where negative behavior mainly occurs and , assuming simply taking the normalization result as the final parameter, then the final generated positive dynamic parameter and the negative dynamic parameter .

[0037] Please refer to Figure 5 , the specific steps for obtaining the reconstructed weight set are as follows: S401: Call the positive dynamic parameter and the negative dynamic parameter. Based on the parameter sequences of three consecutive time windows, align the positive and negative parameters of each window point by point, calculate the correlation coefficient, and at the same time take the mean within the window as the degree of association. Integrate the data of the three windows to generate a window correlation degree set; For example, the parameters of window , , are respectively: Window : ; Window : ; Window : (using the calculation result of the previous step), align the positive and negative parameters within each window point by point and calculate the correlation coefficient between them. Here, the "correlation coefficient" can be defined as a measure of the relationship between the two, such as the difference , or the ratio , or the combination function , meanwhile, "taking the mean value within the window as the correlation degree" may mean taking the mean value if there are multiple data points within the window (such as multiple users), but in this example, there is only one data point in each window Yes, so the "correlation coefficient" is the "correlation degree" of this window. We use the difference as the calculation method for the correlation degree: correlation degree , calculate the correlation degrees of the three windows: , , , integrate the correlation degrees of these three windows to generate a window correlation degree set .

[0038] S402: Extract the adjacent window correlation degrees from the window correlation degree set, calculate the absolute difference between the latter window and the previous window, obtain the differences between the first and second windows and between the second and third windows, compare the two and take the maximum value as the current difference degree. At the same time, numerically compare the current difference degree with a preset difference degree threshold. If it exceeds the threshold, generate a difference degree status flag; Extract from the window correlation degree set the correlation degree values of adjacent windows, calculate the absolute difference between the correlation degree of the latter window and the correlation degree of the previous window, and calculate the absolute difference in the correlation degrees between the first and second windows ; Calculate the absolute difference in the correlation degrees between the second and third windows ; Compare these two differences and , take the maximum value of them as the current difference degree , set a preset difference degree threshold , this threshold is used to judge whether the change in the correlation degree is significant enough. Its setting can be based on the statistical distribution of historical correlation degree changes, such as taking the 85th percentile of the distribution, or setting an empirical value. Suppose the setting is , compare the calculated current difference degree with the preset threshold . Because , the current difference degree exceeds the threshold, so generate a difference degree status flag, marked as "significant change" (or 1).

[0039] S403: According to the difference degree status flag, extract the original priorities of the advertisement content weight set, reallocate the weight values according to preset rules and sort them in descending order to overwrite the original set and generate a reconstructed weight set.

[0040] According to the result of the difference degree status marking, check whether the mark is "significant change" (1). In this example, the mark is 1, so the weight reconstruction operation needs to be performed. First, extract the original priority of the advertisement content weight set (i.e., the original weight value). Assume the original weight set is {Advertisement A: 0.5, Advertisement B: 0.3, Advertisement C: 0.2}. The weight value represents the placement priority, and the larger the value, the higher the priority. Then, re - distribute the weight values according to the preset rules. The rules should clearly stipulate how to adjust the weights when a significant change is detected. For example, the rule can be: "If a significant change is detected, then increase the weight of the advertisement that performs better within the time period with the largest change in association degree (here from to the change, ), assuming that Advertisement A is more strongly associated with the positive parameter during this period) by 20%, and decrease the weight of the advertisement with worse performance (assumed to be Advertisement C) by 20%, and then re - normalize". Specifically execute: increase the weight of Advertisement A , decrease the weight of Advertisement C , the weight of Advertisement B remains unchanged at 0.3, obtaining the new temporary weights {A: 0.6, B: 0.3, C: 0.16}, and the sum is . Perform normalization: the new weight of A , the new weight of B , the new weight of C . Arrange the re - distributed and normalized weights in descending order (already in descending order), obtaining {Advertisement A: 0.566, Advertisement B: 0.283, Advertisement C: 0.151}. Use this new weight set to overwrite the original weight set. If the difference degree status mark of S402 is 0 (not exceeding the threshold), then do not perform reconstruction and keep the original weights unchanged. Finally, generate the reconstructed weight set {Advertisement A: 0.566, Advertisement B: 0.283, Advertisement C: 0.151}.

[0041] Please refer to Figure 6 , the specific steps for obtaining the advertisement resource distribution instruction are as follows: S501: Call the reconstructed weight set, extract the computing resource requirement value and storage resource requirement value of the advertisement placement node, and combine the node load value and delay coefficient in the elastic resource scheduling algorithm of the cloud server. Use the formula: ; Calculate to obtain the node resource allocation parameter value and generate a resource allocation parameter set; Among them, represents the resource allocation parameter of node , is the dynamic reconstruction weight of node in the reconstructed weight set, is the node The computing resource requirement value of For nodes The storage resource requirement value, For nodes The current load value, For nodes The delay coefficient, represents the reconstruction weight association term, represents a delayed associated term; Call the reconstruction weight set {Ad A: 0.566, Ad B: 0.283, Ad C: 0.151}, where Representative Advertising (Related items )’s dynamic reconstruction weights, such as , extract each ad delivery node (Assuming each ad corresponds to a delivery node or service) The required computing resource requirements and storage resource requirements , these values are attributes of the ad itself, for example: Ad A needs Unit computing resources, MB storage resources; Ad B needs Unit computing resources, MB storage resources; Ad C requires Unit computing resources, MB storage resources, combined with the real-time status data of each node provided by the cloud server elastic resource scheduling algorithm, obtain the node Current load value (a normalized value between 0 and 1, indicating the comprehensive utilization of CPU, memory, etc.) and network delay coefficient (For example, the normalized latency indicator, the associated items Refers to delay), assuming the current status of each node is: Node A load ,Delay Node B load ,Delay Node C load ,Delay , using the resource allocation parameter calculation formula: Detailed description of formula parameters: Representative Node The resource allocation parameter value of the resource allocation parameter, the higher the value, the higher the allocation priority; is a node (Associated Ads )’s reconstruction weight; is a node The amount of computing resources required; is a node The amount of storage resources required (Note: directly adding the computing unit and the storage unit may require prior normalization or weighting. For simplicity of calculation, it is assumed that they are already comparable or have been converted to equivalent computing units); is the current load of node ; is the delay coefficient of node (associated delay ); The denominator combines the load and the delay to form a comprehensive node status penalty term. The higher the load or the delay, the larger the denominator, and the lower the allocation parameter . The operation logic is: calculate the basic score according to the dynamic importance of the advertisement ( ) and its total resource requirements ( ), and then adjust (penalize) according to the current load and delay of the node. Nodes with low load and low delay can obtain higher allocation parameters. The benefit of the formula is that it not only considers the dynamic priority and resource requirements of the advertisement content itself, but also combines the running status (load and delay) of the delivery node in real time, making the resource allocation decision more intelligent and efficient, and preferentially allocating important advertisements with high resource requirements to nodes in good condition. Calculate for each advertisement node (assuming has been converted to equivalent computing units, for example ): Node A: ; Node B: ; Node C: ; The operation obtains the resource allocation parameter values of each node, generating a resource allocation parameter set {Node A: 4.478, Node B: 2.404, Node C: 0.396}. This result shows that after comprehensively considering the reconstruction weight, resource requirements, and node status, the resource allocation priority of Node A is the highest. These parameter values will be used to determine which nodes obtain resource adjustment.

[0042] S502: Based on the resource allocation parameter set, numerically compare the node resource allocation parameter values with the priority threshold of the advertisement display queue, eliminate the parameter values lower than the priority threshold, and intercept the parameter value sequence within the upper limit of the queue length according to the time window constraint, generating a node resource adjustment queue; Based on the resource allocation parameter set {Node A: 4.478, Node B: 2.404, Node C: 0.396}, set a priority threshold for the advertisement display queue , only nodes with resource allocation parameter values higher than this threshold will be considered for resource adjustment. The setting of this threshold is based on the minimum quality of service or resource utilization efficiency that the system hopes to guarantee. For example, according to the system capacity and the expected response speed, set , for each node in the set value and for numerical comparison: Node A ( ), Node B ( ), Node C ( ), eliminate the parameter values below the priority threshold, that is, eliminate the parameter values of Node C, and obtain the candidate queue {Node A: 4.478, Node B: 2.404}. Next, intercept the parameter value sequence within the upper limit of the queue length according to the time window constraint. Assume that the upper limit of the number of resource adjustment operations allowed to be executed within the current time window is . Since the length of the candidate queue is 2, which is less than the upper limit of 5, all are retained. If the length of the candidate queue exceeds , then it is necessary to sort by value from high to low, and only take the first ones, and sort the candidate queue in descending order of value (if necessary): {Node A: 4.478, Node B: 2.404}, and generate the final node resource adjustment queue [Node A, Node B].

[0043] S503: Call the node resource adjustment queue, and bind the computing resource demand value and storage resource demand value corresponding to each parameter value in the queue item by item according to the time sequence number of the advertisement display queue to generate an advertisement resource distribution instruction.

[0044] Call the node resource adjustment queue [Node A, Node B]. For each node in the queue, extract its corresponding computing resource demand value and storage resource demand value (information from S501). For example, Node A corresponds to unit computing resources and MB storage resources, and Node B corresponds to unit computing resources and MB storage resources, bind these resource requirement values with the node identifiers item by item according to the chronological number of the advertisement display queue (or adjust the order of the queue itself, such as the order of [Node A, Node B] in this example), to form specific resource scheduling instructions. For example, generate an instruction set: {Instruction 1: {Target Node: 'Node A', Computation Resource Adjustment: 3, Storage Resource Adjustment: 200MB, Priority / Timing: 1}, Instruction 2: {Target Node: 'Node B', Computation Resource Adjustment: 2, Storage Resource Adjustment: 150MB, Priority / Timing: 2}}. This instruction set is the finally generated advertisement resource distribution instruction, which can be handed over to the underlying cloud resource management system for execution.

[0045] An e-commerce promotion system based on cloud computing, the e-commerce promotion system based on cloud computing is used to execute the above-mentioned e-commerce promotion method based on cloud computing. The system includes: A behavior sequence construction module, which is used to obtain user click stream data through cloud log collection nodes, perform data cleaning and timestamp alignment operations on the page jump path, stay duration, and product category labels, input the cleaned data into a hidden Markov model to generate a user behavior sequence, and transfer the user behavior sequence to a mutation point extraction module; A mutation point extraction module, which is used to call the user behavior sequence, perform a sliding window alignment operation on the Euclidean distance and time interval difference of adjacent jump paths based on the dynamic time warping algorithm, extract the fluctuation peak points that exceed the preset threshold three times in a row, generate a mutation point sequence, and transfer the mutation point sequence to a dynamic parameter generation module; A dynamic parameter generation module, which is used to obtain the positive behavior data and negative behavior data in the product interaction log, perform piecewise integral operations on the number of collections, stay duration on the detail page, page skip frequency, and fast sliding frequency based on the mutation point sequence, generate positive dynamic parameters and negative dynamic parameters, and transfer the positive dynamic parameters and negative dynamic parameters to a weight reconstruction module; A weight reconstruction module, which is used to call the positive dynamic parameters and negative dynamic parameters, calculate the correlation degree difference value between the two in three consecutive time windows based on the grey relational analysis method, perform a priority reset operation on the advertisement content weight set when the difference value exceeds the dynamic threshold, generate a reconstructed weight set, and transfer the reconstructed weight set to a resource scheduling mapping module; A resource scheduling mapping module, which is used to call the cloud server elastic resource scheduling algorithm based on the reconstructed weight set, dynamically adjust the allocation ratio of the computation resources and storage resources of the advertisement placement nodes, map the adjusted resource allocation plan to the advertisement display queue, generate an advertisement resource distribution instruction and output it to the cloud execution node.

[0046] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the relevant art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. An e-commerce promotion method based on cloud computing, characterized in that, It includes the following steps: S2: Input the user behavior sequence into the dynamic time warping algorithm, perform sliding window alignment on the Euclidean distance and time interval difference of adjacent jump paths, extract the fluctuation peak points that exceed the threshold three times continuously, and generate a mutation point sequence; S3: Call the positive behavior data and negative behavior data of the commodity interaction log, and perform piecewise integral operations on the number of collections, the duration of staying on the detail page, the page skip frequency, and the fast sliding frequency based on the mutation point sequence to generate positive dynamic parameters and negative dynamic parameters; S4: Input the positive dynamic parameters and the negative dynamic parameters into the grey relational analysis method, calculate the difference in the correlation degree between the two within three consecutive time windows, and when the difference exceeds the preset threshold, perform a priority reset operation on the advertising content weight set to generate a reconstructed weight set; S5: Based on the reconstructed weight set, call the elastic resource scheduling algorithm of the cloud server, allocate the computing resources and storage resources of the advertising placement node, and map them to the advertising display queue to generate an advertising resource distribution instruction.

2. The e-commerce promotion method based on cloud computing according to claim 1, wherein The mutation point sequence includes the fluctuation peak point timestamp, the fluctuation amplitude value, and the adjacent peak point interval. The positive dynamic parameters are specifically the integral value of the number of collections and the integral value of the duration of staying on the detail page. The negative dynamic parameters are specifically the integral value of the page skip frequency and the integral value of the fast sliding frequency. The reconstructed weight set includes the reset weight value, the priority label, and the advertising identifier. The advertising resource distribution instruction specifically refers to the computing node allocation strategy, the storage node allocation strategy, and the advertising display order list.

3. The e-commerce promotion method based on cloud computing according to claim 2, wherein The specific steps for obtaining the mutation point sequence are as follows: S201: Obtain the adjacent jump paths in the user behavior sequence, calculate the spatial difference value of adjacent path coordinate points, extract the timestamp difference of the corresponding paths, and combine the Euclidean distance and the time interval difference into a difference pair in the path order to generate a path difference pair sequence; S20 ； ​ Among them, represents the fluctuation intensity of the th window, represents the variance value of the Euclidean distance between adjacent paths within the window, represents the sum of the absolute values of the differences in the time intervals between adjacent paths within the window, is the cumulative value of the number of fluctuations within the current window, is the fixed length of the sliding window; ​ 4. The e-commerce promotion method based on cloud computing according to claim 3, characterized in that, ​ ​ ​ ； The integral value is obtained by operation, and the integral result of the positive behavior data and the integral result of the negative behavior data are calculated to generate an integral vector; Among them, represents the dynamic integral value of the th type of behavior, represents the number of collections in the th segment, represents the dwell time on the detail page in the th segment, represents the time interval of the mutation point in the th segment, represents the page skip frequency in the th segment, represents the number of slides in the th segment, represents the user activity coefficient in the th segment, represents the balance factor of interaction density and behavior type, represents the total number of segments, represents the start timestamp of the th segment; S303: Separate the positive integral and the negative integral according to the positive and negative signs of the integral vector, normalize the positive integral using the maximum and minimum method, normalize the negative integral using the standard deviation method, and allocate a time weight factor based on the proportion of segment duration to generate positive dynamic parameters and negative dynamic parameters.

5. The e-commerce promotion method based on cloud computing according to claim 4, characterized in that The steps for obtaining the reconstruction weight set are specifically as follows: S401: Calling the positive dynamic parameter and the negative dynamic parameter, based on the parameter sequence of three consecutive time windows, aligning the positive and negative parameters of each window point by point, and calculating the correlation coefficient. At the same time, taking the mean value within the window as the correlation degree, integrating the three window data, and generating a window correlation degree set; S402: extracting adjacent window correlations from the window correlation set, calculating the absolute difference between the subsequent window and the preceding window, obtaining the difference between the first and second windows, and the difference between the second and third windows, comparing the two and taking the maximum value as the current difference, and comparing the current difference with a preset difference threshold. If the difference exceeds the threshold, generating a difference status flag; S403: extracting the original priority of the advertising content weight set according to the difference status mark, reallocating the weight values according to a preset rule and arranging them in descending order, overwriting the original set, and generating a reconstructed weight set.

6. The e-commerce promotion method based on cloud computing according to claim 5, wherein, The steps for obtaining the advertising resource distribution instruction are specifically as follows: S501: Call the reconstruction weight set to extract the computing resource demand value and storage resource demand value of the advertisement delivery node, and combine the node load value and delay coefficient in the cloud server elastic resource scheduling algorithm to use the formula: ； Obtain node resource allocation parameter values through calculation and generate a resource allocation parameter set; Among them, represents the resource allocation parameter of the node , is the dynamic reconstruction weight of the node in the reconstruction weight set, is the computing resource requirement value of the node , is the storage resource requirement value of the node , is the current load value of the node , is the delay coefficient of the node , represents the reconstruction weight correlation item, represents the delay correlation item; S502: Based on the resource allocation parameter set, compare the node resource allocation parameter value with the priority threshold of the advertisement display queue, eliminate parameter values below the priority threshold, intercept the parameter value sequence within the queue length upper limit according to the time window constraint, and generate a node resource adjustment queue; S503: calling the node resource adjustment queue, binding the computing resource demand value and the storage resource demand value corresponding to each parameter value in the queue one by one according to the time sequence number of the advertisement display queue, and generating an advertisement resource distribution instruction.

7. The e-commerce promotion method based on cloud computing according to claim 6, characterized in that The method further comprises: S1: Obtain user clickstream data through cloud log collection nodes, perform cleaning and alignment operations on page jump paths, dwell time, and product category labels, and construct user behavior sequences based on the Hidden Markov Model; The user behavior sequence specifically includes a jump timestamp, a subject code, and a page level value.

8. The e-commerce promotion method based on cloud computing according to claim 7, characterized in that The steps for obtaining the user behavior sequence are specifically as follows: S101: Collecting raw user clickstream data from cloud logs, extracting page jump paths, dwell time, and product category label fields, filtering out missing path nodes and abnormal dwell time values, matching product category labels with page paths, and generating standardized clickstream data; S102: Based on the standardized clickstream data, align the timestamps of the page jump paths and product category labels, unify the time precision of different terminals, accumulate the dwell time segment by segment according to the path nodes, extract the continuous path, duration, and label combinations under the same user ID, and generate a set of behavior trajectory parameters. S103: Invoke the path jump frequency, average dwell time, and label correlation degree in the set of behavior trajectory parameters, calculate the state transition probability matrix and the dwell time distribution density, map the path nodes to the state variables of the hidden Markov model, establish the correspondence between the state transition and the observation sequence, and generate a user behavior sequence model.

9. An e-commerce promotion system based on cloud computing, characterized in that, The system is used to implement the cloud computing-based e-commerce promotion method described in any one of claims 1-8. The system includes: A behavior sequence construction module, configured to obtain user clickstream data through a cloud log collection node, perform data cleaning and timestamp alignment operations on the page jump path, dwell time, and product category label, input the cleaned data into a hidden Markov model to generate a user behavior sequence, and transfer the user behavior sequence to a mutation point extraction module. A mutation point extraction module, configured to invoke the user behavior sequence, perform a sliding window alignment operation on the Euclidean distance and time interval difference between adjacent jump paths based on the dynamic time warping algorithm, extract the fluctuation peak points that exceed the preset threshold three times continuously, generate a mutation point sequence, and transfer the mutation point sequence to a dynamic parameter generation module. A dynamic parameter generation module, configured to obtain the positive behavior data and negative behavior data in the product interaction log, perform a segmented integration operation on the number of favorites, dwell time on the detail page, page skip frequency, and fast sliding frequency based on the mutation point sequence, generate positive dynamic parameters and negative dynamic parameters, and transfer the positive dynamic parameters and the negative dynamic parameters to a weight reconstruction module. A weight reconstruction module, configured to invoke the positive dynamic parameters and the negative dynamic parameters, calculate the difference value of the correlation degree between the two in three consecutive time windows based on the grey relational analysis method, perform a priority reset operation on the advertising content weight set when the difference value exceeds the dynamic threshold, generate a reconstructed weight set, and transfer the reconstructed weight set to a resource scheduling mapping module. A resource scheduling mapping module, configured to call the cloud server elastic resource scheduling algorithm based on the reconstructed weight set, dynamically adjust the allocation ratio of the computing resources and storage resources of the advertising placement node, map the adjusted resource allocation plan to the advertising display queue, generate an advertising resource distribution instruction, and output it to the cloud execution node.

Images