Cross-border e-commerce market demand prediction system based on big data

Abstract

The invention relates to the technical field of market demand analysis, in particular to a big data-based cross-border e-commerce market demand prediction system, which comprises a culture trend identification module, a sudden demand evaluation module and a demand prediction output module, wherein the culture trend identification module is used for capturing text data of sudden popular culture in a short period through a multi-dimensional data source; the sudden demand evaluation module is combined with sudden popular culture to analyze and quantify demand fluctuation for hot commodities in a short period; and the demand prediction output module provides promotion and marketing activity suggestions of hot commodities and inventory adjustment strategies according to the demand change prediction result. According to the method, the dependence of a traditional demand prediction method on a long-term trend is broken through, the accuracy of short-term demand prediction is remarkably improved, and a cross-border e-commerce platform is helped to respond to market changes in time.

Description

Technical Field

[0001] The present invention relates to the technical field of market demand analysis, and in particular to a cross-border e-commerce market demand forecasting system based on big data. Background Art

[0002] In recent years, the cross-border e-commerce market has experienced rapid development. In particular, the trend of globalization and the continuous advancement of Internet technology have made cross-border e-commerce an important part of global retail. In order to stand out in the fierce market competition, cross-border e-commerce platforms and merchants are increasingly relying on accurate market demand forecasting systems, aiming to improve product sales efficiency and inventory management through data-driven decision-making.

[0003] At present, market demand forecasting technology mainly relies on historical sales data and seasonal trend analysis. These traditional methods generally use time series analysis, regression analysis and other techniques to predict future demand changes by analyzing past sales data. Although these methods perform well in long-term demand forecasting, they often have great limitations when dealing with sudden demand changes such as short-term market fluctuations, unexpected events and popular trends.

[0004] Unable to respond to the impact of short-term trends and emergencies in a timely manner: Most existing demand forecasting methods are based on historical data and long-term trends, and cannot quickly respond to the impact of social media, sudden popular cultural events, sudden social hot spots, etc. on market demand. Therefore, when facing a sudden burst of demand in the short term, cross-border e-commerce platforms are often unable to make timely adjustments, resulting in excess or shortage of inventory, affecting sales and user experience. Summary of the invention

[0005] The present invention provides a cross-border e-commerce market demand forecasting system based on big data.

[0006] The cross-border e-commerce market demand forecasting system based on big data includes a cultural trend identification module, a sudden demand assessment module, and a demand forecast output module, among which: The cultural trend identification module is used to capture text data of sudden popular culture in a short period of time through a multi-dimensional data source, wherein the multi-dimensional data source includes a social media platform and an e-commerce platform, and the text data includes: User comments and search data on hot products on social media platforms, including hot products, interaction frequency, and hot topics posted by users on Douyin, Xiaohongshu, and Bilibili; Real-time user comments and search data on hot commodities on e-commerce platforms, including sales changes of hot commodities related to sudden topics on Taobao and JD.com; The sudden demand assessment module combines sudden popular culture to analyze and quantify the demand fluctuations for hot commodities in the short term, including: Sentiment analysis based on the BERT model analyzes the sentiment of text data on hot commodities in real time, obtains sentiment scores, and quantifies the demand driving coefficient of sentiment scores on hot commodities; Based on the demand push coefficient, a sudden demand forecasting model is constructed to forecast demand changes for hot commodities in the short term to adapt to the rapidly changing short-term trends; The demand forecast output module provides promotion of hot commodities, marketing activity suggestions and inventory adjustment strategies based on the demand change forecast results.

[0007] Optionally, the multi-dimensional data source is obtained by providing public API interfaces through major social media platforms and open API interfaces provided by e-commerce platforms, allowing developers to obtain product data, user reviews, sales and search information.

[0008] Optionally, obtaining the sentiment score specifically includes: Sentiment classification: Use the BERT model to classify the collected text data into sentiments, distinguishing positive, negative, and neutral sentiments; Sentiment score: Based on the sentiment classification results, each comment or search information is assigned a separate sentiment score (for example, a value from -1 to +1, indicating the intensity of negative to positive sentiment), and an overall comprehensive product sentiment score is generated through weighted averaging.

[0009] Optionally, the BERT model performs sentiment classification on the collected text data. The input text data is converted into sentiment labels, including positive, negative, and neutral, through the trained BERT model. The sentiment classification result output by the BERT model is : ,in, Expressing negative emotions, Expressing neutral emotions, Express positive emotions; According to the sentiment classification results, each user comment or search data is assigned a separate sentiment score. Calculated as: ,in, It is Sentiment ratings of comments or searches, is the sentiment classification result of the user's comments or search data. is a constant representing the weight of the sentiment score ( ), in order to quantify the intensity of sentiment, sentiment scores are normalized from -1 to +1, representing the intensity from completely negative to completely positive; Generate a comprehensive product sentiment score by taking a weighted average of the sentiment scores of all user comments or search data .

[0010] Optionally, the comprehensive commodity sentiment score , calculated as: ,in, is the comprehensive commodity sentiment score of hot commodities. is the total number of reviews or search information collected, It is The weight of a comment or search information is proportional to the influence of the comment (such as the number of likes, shares, etc.).

[0011] Optionally, the quantitative sentiment score driving coefficient on the demand for hot commodities specifically includes: Through regression analysis, combined with historical sales data and sentiment scores , calculate the demand driving coefficient of sentiment score on hot products .

[0012] Optionally, in the regression analysis, historical sales data for , historical sentiment scores for , the basic model of regression analysis is the linear regression model, which is expressed as: ,in, It is Sales data at the moment, It is The emotional score of the moment, is the coefficient of sentiment score on demand promotion, that is, demand promotion coefficient, is the error term.

[0013] Optionally, the regression analysis also includes fitting the regression equation by the least square method to obtain the demand driving coefficient The best estimate of : ,in, is the estimated value of the sentiment score's demand-driving coefficient, and are the means of sentiment scores and sales data respectively.

[0014] Optionally, the sudden demand prediction model is expressed as: ,in, Hot Products In the future The demand forecast, Hot Products At the present moment The baseline demand, Hot Products The demand driving coefficient indicates the impact of sentiment scores on the demand for hot commodities in the short term. is the forecast time span.

[0015] Beneficial effects of the present invention: The present invention combines multi-dimensional data sources, such as instant user comments, interaction frequency, hot topics and other information on social media platforms and e-commerce platforms, to accurately capture sudden changes in popular culture in the short term, and then identify hot commodities. It uses natural language processing technology to perform sentiment analysis, monitor the fluctuations of consumer emotions in real time, quantify them into sentiment scores, and combine them with historical data through a demand-driven coefficient model. It can effectively predict short-term demand changes, breaking through the reliance of traditional demand forecasting methods on long-term trends, significantly improving the accuracy of short-term demand forecasting, and helping cross-border e-commerce platforms to respond to market changes in a timely manner. The present invention can automatically generate sales promotion, marketing activity suggestions and inventory adjustment strategies based on the sudden demand forecast results. By providing accurate promotion plans based on the demand forecast results and dynamically adjusting the inventory in combination with demand changes, the system can effectively improve the market exposure and sales volume of hot commodities. At the same time, with the support of inventory adjustment strategies, the platform can avoid inventory backlogs caused by excess demand or out-of-stock problems caused by low demand, improve the efficiency of resource use, reduce operating costs, and enhance the adaptability and profitability of cross-border e-commerce platforms in the fiercely competitive market. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 is a system logic block diagram of an embodiment of the present invention; Figure 2 Schematic diagram of system function modules according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0019] It should be noted that the references to "one embodiment", "embodiment", "exemplary embodiments", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0020] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0021] like Figure 1-Figure 2 As shown in the figure, the cross-border e-commerce market demand forecasting system based on big data includes a cultural trend identification module, a sudden demand assessment module and a demand forecasting output module, among which: The cultural trend identification module is used to capture the text data of sudden popular culture in the short term through multi-dimensional data sources, including social media platforms and e-commerce platforms. The text data includes: User comments and search data on hot products on social media platforms, including hot products, interaction frequency, and hot topics posted by users on Douyin, Xiaohongshu, and Bilibili; Real-time user comments and search data on hot commodities on e-commerce platforms, including sales changes of hot commodities related to sudden topics on Taobao and JD.com; The sudden demand assessment module combines sudden popular culture to analyze and quantify the demand fluctuations for hot commodities in the short term, including: Sentiment analysis based on the BERT model analyzes the sentiment of text data on hot commodities in real time, obtains sentiment scores, and quantifies the demand driving coefficient of sentiment scores on hot commodities; Based on the demand push coefficient, a sudden demand forecasting model is constructed to forecast demand changes for hot commodities in the short term to adapt to the rapidly changing short-term trends; The demand forecast output module provides promotion of hot commodities, marketing activity suggestions and inventory adjustment strategies based on the demand change forecast results.

[0022] Sales recommendations: Based on the demand change forecast results, the demand forecast output module provides sales recommendations for hot products. Specific methods include: When demand forecasts show that demand for a product will grow rapidly, it is recommended to increase product exposure, advertising, or recommendations to take advantage of the rising demand trend by increasing sales efforts; When the demand forecast results show that the demand for a certain product will decrease, it is recommended to reduce the promotion efforts for that product to avoid wasting resources and focus on promoting other products with higher demand.

[0023] Marketing campaign suggestions: Based on the demand change forecast results, the demand forecast output module provides customized marketing campaign suggestions for hot products. Specific methods include: High-demand products: When demand forecasts show that the demand for a product will increase significantly, it is recommended to launch emergency promotions such as limited-time discounts and flash sales to drive consumption by creating a sense of urgency; Low-demand products: When demand forecast results show that demand for a certain product is decreasing or fluctuating, it is recommended to stimulate demand growth by increasing marketing efforts or through bundling sales.

[0024] Inventory adjustment strategy: Based on the demand change forecast results, the demand forecast output module provides inventory adjustment strategies for hot commodities. Specific methods include: When demand growth is predicted: When the demand forecast results show that the demand for a certain product will increase significantly, it is recommended to prepare in advance and increase the inventory of the product to avoid sales losses due to insufficient supply; When demand is predicted to decrease: When the demand forecast results show that the demand for a certain product will decrease, it is recommended to reduce the inventory of the product and conduct inventory clearance, such as accelerating inventory consumption through discounts or promotions; When demand fluctuates: For commodities with large demand fluctuations, it is recommended to flexibly adjust inventory based on the forecast fluctuations in the short term to maximize inventory utilization and reduce excess inventory.

[0025] Multidimensional data sources are obtained through public API interfaces provided by major social media platforms and open API interfaces provided by e-commerce platforms, allowing developers to obtain product data, user reviews, sales and search information.

[0026] Obtaining sentiment scores specifically includes: Sentiment classification: Use the BERT model to classify the collected text data into sentiments, distinguishing positive, negative, and neutral sentiments; Sentiment score: Based on the sentiment classification results, each comment or search information is assigned a separate sentiment score (for example, a value from -1 to +1, indicating the intensity of negative to positive sentiment), and an overall comprehensive product sentiment score is generated through weighted averaging.

[0027] The BERT model performs sentiment classification on the collected text data. Through the trained BERT model, the input text data is converted into sentiment labels, including positive, negative, and neutral. The sentiment classification results output by the BERT model are : ,in, Expressing negative emotions, Expressing neutral emotions, Express positive emotions; According to the sentiment classification results, each user comment or search data is assigned a separate sentiment score. Calculated as: ,in, It is Sentiment ratings of comments or searches, is the sentiment classification result of the user's comments or search data. is a constant representing the weight of the sentiment score ( ), in order to quantify the intensity of sentiment, sentiment scores are normalized from -1 to +1, representing the intensity from completely negative to completely positive; Generate a comprehensive product sentiment score by taking a weighted average of the sentiment scores of all user comments or search data .

[0028] The specific steps of BERT model sentiment classification are as follows: 1: Data preprocessing: Collect comments and search information text data from social media and e-commerce platforms, clean the data, remove irrelevant information such as punctuation and special characters; split the text data into words or sub-words. BERT uses sub-word-level tokenization and adopts the WordPiece algorithm. For example, for Chinese text, the text is first broken down into single words or short word fragments. For English text, BERT will split words into sub-word units; in order to adapt to the input format of the BERT model, special tags need to be added to each text, such as [CLS] (text start tag) and [SEP] (text separation tag). These tags help the model understand the beginning and end of the text.

[0029] 2: Text input to BERT model: Each input text is converted into a pair of tag IDs, and the text is converted into a digital form using a vocabulary. The input of the BERT model consists of three parts: Token IDs: Convert text into digital IDs through a vocabulary; Attention Mask: used to specify which tokens are actual content and which are filler content; Segment IDs: mark sentence A and sentence B in the input text (for sentence pair tasks); BERT Embedding Layer: The input tokens are converted into vector representations through BERT's embedding layer. These vectors capture the contextual information of the text and can understand the polysemy and contextual relationships of words.

[0030] 3: Sentiment classification layer: Encoding through Transformer layers: The core of the BERT model is a multi-layer Transformer network, which is used to model the context of text. The Transformer structure captures the relationship between words through the self-attention mechanism.

[0031] Extract the output of the [CLS] tag: BERT's [CLS] tag output represents the summary information of the entire text and is usually used for text classification tasks. By taking out the output vector of the [CLS] tag (that is, the last layer output of the BERT model), it represents the contextual information of the entire text.

[0032] Sentiment classification layer: The output vector of the [CLS] tag is input into a fully connected layer and a softmax activation function to generate the output of the sentiment classification. Through the trained weights, the model will output three categories: positive, negative, and neutral.

[0033] 4: Sentiment classification and scoring: Sentiment classification: Use the Softmax activation function to get the probability distribution of the sentiment category of each comment: ,in, is the logit value for each category, indicating the predicted probability of that category, is the base of natural logarithms, The index representing the category; Emotional category determination: Determine the emotional category of the text based on the maximum probability value output by Softmax: If the probability of positive sentiment is the highest, it is classified as positive; If the probability of negative sentiment is the highest, it is classified as negative; If the probability of neutral sentiment is the highest, then classify it as neutral.

[0034] Comprehensive product sentiment score , calculated as: ,in, is the comprehensive sentiment score of hot commodities. is the total number of reviews or search information collected, It is The weight of a comment or search information is proportional to the influence of the comment (such as the number of likes, shares, etc.).

[0035] The specific factors driving the demand for hot commodities through quantitative sentiment scores include: Through regression analysis, combined with historical sales data and sentiment scores , calculate the demand driving coefficient of sentiment score on hot products .

[0036] In regression analysis, historical sales data for , historical sentiment scores for , the basic model of regression analysis is the linear regression model, which is expressed as: ,in, It is Sales data at the moment, It is The emotional score of the moment, is the coefficient of sentiment score on demand promotion, that is, demand promotion coefficient, is the error term.

[0037] Regression analysis also includes fitting the regression equation through the least squares method to obtain the demand driving coefficient The best estimate of : ,in, is the estimated value of the sentiment score's demand-driving coefficient, and are the means of sentiment scores and sales data respectively.

[0038] The sudden demand forecasting model is expressed as: ,in, Hot Products In the future The demand forecast, Hot Products At the present moment Baseline demand (i.e. historical average demand data, for commodities In the absence of sudden changes in demand), Hot Products The demand driving coefficient indicates the impact of sentiment scores on the demand for hot commodities in the short term. is the forecast time span.

[0039] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0040] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. The cross-border e-commerce market demand forecasting system based on big data is characterized by: It includes cultural trend identification module, sudden demand assessment module and demand forecast output module, among which: The cultural trend identification module is used to capture text data of sudden popular culture in a short period of time through a multi-dimensional data source, wherein the multi-dimensional data source includes a social media platform and an e-commerce platform, and the text data includes: User comments and search data on popular products on social media platforms; Real-time user reviews and search data on hot products on e-commerce platforms; The sudden demand assessment module combines sudden popular culture to analyze and quantify the demand fluctuations for hot commodities in the short term, including: Sentiment analysis based on the BERT model analyzes the sentiment of text data on hot commodities in real time, obtains sentiment scores, and quantifies the demand driving coefficient of sentiment scores on hot commodities; Based on the demand driving coefficient, a sudden demand forecasting model is constructed to predict demand changes for hot commodities in the short term; The demand forecast output module provides promotion of hot commodities, marketing activity suggestions and inventory adjustment strategies based on the demand change forecast results.

2. The cross-border e-commerce market demand forecasting system based on big data according to claim 1 is characterized in that: The multi-dimensional data source is obtained through the public API interfaces provided by major social media platforms and the open API interfaces provided by e-commerce platforms, allowing developers to obtain product data, user reviews, sales and search information.

3. The cross-border e-commerce market demand forecasting system based on big data according to claim 1 is characterized in that: The obtaining of the sentiment score specifically includes: Sentiment classification: Use the BERT model to classify the collected text data into sentiments, distinguishing positive, negative, and neutral sentiments; Sentiment score: Based on the sentiment classification results, each comment or search information is assigned a separate sentiment score, and the overall comprehensive product sentiment score is generated by weighted average.

4. The cross-border e-commerce market demand forecasting system based on big data according to claim 3 is characterized in that: The BERT model performs sentiment classification on the collected text data. Through the trained BERT model, the input text data is converted into sentiment labels, including positive, negative, and neutral. The sentiment classification result output by the BERT model is : ,in, Expressing negative emotions, Expressing neutral emotions, Express positive emotions; According to the sentiment classification results, each user comment or search data is assigned a separate sentiment score. Calculated as: ,in, It is Sentiment ratings of comments or searches, is the sentiment classification result of the user's comments or search data. is a constant representing the weight of the sentiment score, which is normalized from -1 to +1, representing the intensity from completely negative to completely positive; Generate a comprehensive product sentiment score by taking a weighted average of the sentiment scores of all user comments or search data .

5. The cross-border e-commerce market demand forecasting system based on big data according to claim 4 is characterized in that: The comprehensive product sentiment score , calculated as: ,in, is the comprehensive sentiment score of hot commodities. is the total number of reviews or search information collected, It is The weight of a comment or search information.

6. The cross-border e-commerce market demand forecasting system based on big data according to claim 5 is characterized in that: The quantitative sentiment score driving coefficient of the demand for hot commodities specifically includes: Through regression analysis, combined with historical sales data and sentiment scores , calculate the demand driving coefficient of sentiment score on hot products .

7. The cross-border e-commerce market demand forecasting system based on big data according to claim 6 is characterized in that: In the regression analysis, historical sales data for , historical sentiment scores for , the basic model of regression analysis is the linear regression model, which is expressed as: ,in, It is Sales data at the moment, It is The emotional score of the moment, is the coefficient of sentiment score on demand promotion, that is, demand promotion coefficient, is the error term.

8. The cross-border e-commerce market demand forecasting system based on big data according to claim 7 is characterized in that: The regression analysis also includes fitting the regression equation by the least square method to obtain the demand driving coefficient The best estimate of : ,in, is the estimated value of the sentiment score's demand-driving coefficient, and are the means of sentiment scores and sales data respectively.

9. The cross-border e-commerce market demand forecasting system based on big data according to claim 8 is characterized in that: The sudden demand prediction model is expressed as: ,in, Hot Products In the future The demand forecast, Hot Products At the present moment The baseline demand, Hot Products The demand driving coefficient indicates the impact of sentiment scores on the demand for hot commodities in the short term. is the forecast time span.

Images