Commodity sales data intelligent analysis system and method

Abstract

The invention relates to the field of intelligent analysis, and particularly discloses a commodity sales data intelligent analysis system and method.The commodity sales data intelligent analysis method comprises the steps that firstly, user purchase tendency text data collected by crawlers and Huzhou pen sales condition text data collected from a database in a preset time period are obtained; performing feature extraction on the user purchase tendency text data to obtain a text understanding feature vector, performing normalization processing and cascading on the Huzhou-stylus sales condition text data in the predetermined time period through a semantic understanding model to obtain a Huzhou-stylus sales normalization feature vector, and finally, performing classification on the Huzhou-stylus sales normalization feature vector. And the text understanding feature vector and the Huzhou pen sales normalization feature vector are fused and optimized, and a classifier is used to predict whether the sales amount of the next month is increased or decreased compared with the sales amount of the current month, so that the sales strategy and resource configuration are adjusted in time, and the enterprise is helped to make a reasonable sales strategy and decision.

Description

Technical Field

[0001] This application relates to the field of intelligent analysis, and more specifically, to a system and method for intelligent analysis of commodity sales data. Background Technology

[0002] Sales data for Huzhou brushes may be affected by quarterly variations. As one of the Four Treasures of the Study, the sales of Huzhou brushes can be influenced by factors such as seasonal demand, promotional activities, and the quarterly economic environment, leading to fluctuations in sales data between different quarters. For example, sales of Huzhou brushes may increase during the back-to-school season, exam season, or shopping season; while sales may be relatively low at other times.

[0003] Traditional sales forecasting methods typically involve collecting historical data and then using that data to predict future sales figures and product activity levels. However, this approach to historical data analysis doesn't consider consumer purchasing tendencies. Consumer buying behavior is influenced by various factors, including personal preferences, market trends, promotional activities, and product characteristics, which may not be fully captured by historical sales data.

[0004] Therefore, there is a need for an intelligent analysis system and method for commodity sales data. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide an intelligent analysis system and method for commodity sales data. First, it acquires user purchase tendency text data collected by a web crawler and text data on the sales of calligraphy brushes over a predetermined time period collected from a database. Then, it extracts features from the user purchase tendency text data to obtain a text understanding feature vector. Next, it normalizes and concatenates the calligraphy brush sales data over the predetermined time period using a semantic understanding model to obtain a normalized feature vector for calligraphy brush sales. Finally, it fuses the text understanding feature vector and the normalized feature vector for calligraphy brush sales, optimizes the results, and uses a classifier to predict whether next month's sales will increase or decrease compared to this month's sales, thereby allowing for timely adjustments to sales strategies and resource allocation, ultimately helping enterprises make reasonable sales strategies and decisions.

[0006] According to one aspect of this application, a smart analysis system for commodity sales data is provided, comprising:

[0007] The text data acquisition module is used to acquire user purchase tendency text data collected by the crawler and Huzhou brush sales text data collected from the database for a predetermined time period. The Huzhou brush sales text data includes sales amount, sales quantity, promotional activities and sales time.

[0008] The text data feature extraction module is used to extract text understanding feature vectors and normalized feature vectors of Huzhou brush sales from the user purchase tendency text data collected by the crawler and the Huzhou brush sales data for a predetermined time period collected from the database.

[0009] The sales amount prediction module is used to predict whether the sales amount in the next month will increase or decrease compared to the sales amount in this month, based on the text understanding feature vector and the normalized feature vector of the Huzhou brush sales.

[0010] According to another aspect of this application, a method for intelligent analysis of commodity sales data is provided, comprising:

[0011] The system acquires text data on user purchasing preferences collected by a web crawler and text data on the sales of Huzhou brushes for a predetermined time period collected from a database. The text data on the sales of Huzhou brushes includes sales amount, sales quantity, promotional activities, and sales time.

[0012] Extract text understanding feature vectors and normalized feature vectors of Huzhou brush sales from the user purchase tendency text data collected by the crawler and the Huzhou brush sales data for a predetermined time period collected from the database;

[0013] Based on the text understanding feature vector and the normalized feature vector of Huzhou brush sales, predict whether the sales amount next month will increase or decrease compared to the sales amount this month.

[0014] Compared with existing technologies, this application provides a smart analysis system and method for commodity sales data. First, it acquires textual data on user purchasing tendencies collected by web crawlers and textual data on the sales of calligraphy brushes over a predetermined time period collected from a database. Then, it extracts features from the user purchasing tendency textual data to obtain a text understanding feature vector. Next, it normalizes and concatenates the textual data on the sales of calligraphy brushes over the predetermined time period using a semantic understanding model to obtain a normalized feature vector for calligraphy brush sales. Finally, it fuses the text understanding feature vector and the normalized feature vector for calligraphy brush sales, optimizes the results, and uses a classifier to predict whether the sales amount next month will increase or decrease compared to this month's sales amount. This allows for timely adjustments to sales strategies and resource allocation, thereby helping enterprises make reasonable sales strategies and decisions. Attached Figure Description

[0015] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 This is a block diagram of a commodity sales data intelligent analysis system according to an embodiment of this application.

[0017] Figure 2 This is a block diagram of the text data feature extraction module in the intelligent analysis system for commodity sales data according to an embodiment of this application.

[0018] Figure 3 This is a block diagram of the purchase tendency text feature extraction unit in the intelligent analysis system for commodity sales data according to an embodiment of this application.

[0019] Figure 4 This is a block diagram of the sales amount prediction module in the intelligent analysis system for commodity sales data according to an embodiment of this application.

[0020] Figure 5 This is a schematic diagram of the architecture of a smart analysis system for commodity sales data according to an embodiment of this application.

[0021] Figure 6 This is a flowchart of a product sales data intelligent analysis method according to an embodiment of this application.

[0022] Figure 7 This is a block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0023] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0024] Exemplary System

[0025] Figure 1 This is a block diagram of a product sales data intelligent analysis system according to an embodiment of this application. Figure 1 As shown, the intelligent analysis system 100 for commodity sales data according to an embodiment of this application includes: a text data acquisition module 110, used to acquire user purchase tendency text data collected by a web crawler and Huzhou brush sales data for a predetermined time period collected from a database, wherein the Huzhou brush sales data includes sales amount, sales quantity, promotional activities, and sales time; a text data feature extraction module 120, used to extract text understanding feature vectors and Huzhou brush sales normalized feature vectors from the user purchase tendency text data collected by the web crawler and the Huzhou brush sales data for a predetermined time period collected from the database; and a sales amount prediction module 130, used to predict whether the sales amount next month will increase or decrease compared to the sales amount this month based on the text understanding feature vectors and the Huzhou brush sales normalized feature vectors.

[0026] In the aforementioned intelligent analysis system 100 for commodity sales data, the text data acquisition module 110 is used to acquire user purchase tendency text data collected by a web crawler and text data on the sales of Huzhou brushes within a predetermined time period collected from a database. The Huzhou brush sales text data includes sales amount, sales quantity, promotional activities, and sales time. It should be understood that user purchase tendency text data refers to text information related to purchase intentions published by users on different platforms or channels, such as comments, articles, and requests. By acquiring this text data, it is possible to understand users' purchase preferences, opinions, and needs, thereby better understanding user behavior and market trends. Huzhou brush sales text data refers to text information related to Huzhou brush sales obtained from the database, including sales amount, sales quantity, promotional activities, and sales time. This data can provide information about the actual situation and trends of Huzhou brush sales, used to analyze sales performance, evaluate promotional effects, and predict future sales trends. By acquiring and combining these two types of text data, the intelligent analysis system for commodity sales data can perform text feature extraction and semantic understanding, thereby obtaining a text understanding feature vector and a Huzhou brush sales normalized feature vector. These feature vectors can be used to predict whether sales will increase or decrease in the following month, helping companies to formulate sales strategies and optimize decisions.

[0027] In the aforementioned intelligent analysis system 100 for commodity sales data, the text data feature extraction module 120 is used to extract text understanding feature vectors and normalized feature vectors for brush sales from the user purchase tendency text data collected by web crawlers and the brush sales data for a predetermined time period collected from the database. It should be understood that the extraction of text understanding feature vectors involves the analysis and understanding of user purchase tendency text data. By using natural language processing technology, text data can be transformed into computer-understandable feature vectors, including word vectors, sentence vectors, or other representations. These feature vectors can capture semantic, sentiment, and thematic information in the text, thereby helping to analyze user purchase tendencies. The extraction of normalized feature vectors for brush sales involves the normalization processing of sales data. Sales data typically includes information such as sales amount, sales quantity, promotional activities, and sales time. To compare and analyze these different types of data, normalization processing is required, transforming them into a unified feature vector representation, which can eliminate scale differences between different data. By extracting text understanding feature vectors and normalized feature vectors for brush sales, different types of data can be represented uniformly.

[0028] Figure 2 This is a block diagram of the text data feature extraction module in the intelligent analysis system for commodity sales data according to an embodiment of this application. Figure 2As shown in a specific embodiment of this application, the text data feature extraction module 120 includes: a purchase tendency text feature extraction unit 121, used to extract features from the user purchase tendency text data collected by the crawler to obtain the text understanding feature vector; a sales situation semantic understanding unit 122, used to process the Huzhou brush sales situation text data collected from the database for a predetermined time period through a deep learning-based semantic understanding model to obtain the multiple Huzhou brush sales situation feature vectors; and a sales situation data processing unit 123, used to process the multiple Huzhou brush sales situation feature vectors to obtain the Huzhou brush sales normalized feature vector. It should be understood that text data usually exists in the form of natural language, containing a large number of words and sentences. In order to quantify and calculate the text, it needs to be converted into a feature vector form that computers can understand and process. By extracting features from the user purchase tendency text data, the text data can be converted into a text understanding feature vector, thereby realizing the quantification and analysis of user purchase tendency. Common feature extraction methods include the Bag-of-Words model and TF-IDF (Term Frequency-Inverse Document Frequency). Feature extraction is the process of converting text data into feature vectors. These methods can extract feature information such as keywords, term frequencies, and document frequencies from text and encode them into vector form.

[0029] Furthermore, deep learning-based semantic understanding models can automatically extract semantic information and contextual relationships from text by learning from large amounts of text data. Such models can better understand the meaning of text and capture its hidden features and relationships. By inputting text data on Huzhou brush sales into a deep learning-based semantic understanding model, multiple feature vectors can be obtained. Each feature vector may represent different semantic information, such as sales revenue, sales volume, and promotional activities. These feature vectors can provide a more comprehensive and finer-grained description and analysis of sales, better capturing multiple aspects and details of sales, thus providing a more comprehensive information foundation and supporting more accurate sales analysis, prediction, and decision-making. Specifically, the text data on Huzhou brush sales over a predetermined time period collected from the database is segmented to obtain a word sequence; the embedding layer of the deep learning-based semantic understanding model is used to map each word in the word sequence into a word embedding vector to obtain a sequence of word embedding vectors; the BERT model based on a converter of the deep learning-based semantic understanding model is used to perform global contextual semantic encoding on the sequence of embedding vectors to obtain multiple Huzhou brush sales feature vectors.

[0030] Furthermore, to unify the numerical ranges of different features and eliminate dimensional differences, enabling comparison and analysis on the same scale, multiple feature vectors representing Huzhou brush sales are processed to obtain a normalized feature vector. Normalization helps eliminate the influence of dimensions in the data, making the impact of different features on the model more balanced. In Huzhou brush sales, there may be multiple features, such as sales quantity, sales revenue, and sales growth rate. These features may have different numerical ranges and units. Directly using them for analysis and modeling may lead to some features having an excessively large impact on the results, while ignoring the importance of other features. By normalizing multiple feature vectors representing Huzhou brush sales, the numerical ranges of different features can be mapped to a unified interval, such as [0,1] or [-1,1]. This ensures the comparability of values ​​between different features and better reflects their relative importance in sales. Specifically, common normalization methods include min-max normalization and standardization.

[0031] Figure 3 This is a block diagram of a purchase tendency text feature extraction unit in a product sales data intelligent analysis system according to an embodiment of this application. Figure 3As shown in a specific embodiment of this application, the purchase tendency text feature extraction unit 121 includes: a purchase tendency semantic understanding subunit 1211, used to process the user purchase tendency text data collected by the crawler through a context-based encoder model containing an embedding layer to obtain multiple purchase tendency semantic feature vectors; a purchase tendency two-dimensional feature encoding subunit 1212, used to perform two-dimensional feature encoding on the multiple purchase tendency semantic feature vectors to obtain a purchase tendency associated semantic feature vector; a purchase tendency one-dimensional feature encoding subunit 1213, used to perform one-dimensional feature encoding on the multiple purchase tendency semantic feature vectors to obtain a purchase tendency phrase-level semantic feature vector; and a semantic understanding association subunit 1214, used to concatenate the purchase tendency associated semantic feature vector and the purchase tendency phrase-level semantic feature vector to obtain the text understanding feature vector. It should be understood that a context-based encoder model can learn from a large amount of text data, transforming text into a continuous vector representation and capturing its semantic information and contextual relationships. Such a model can better understand the meaning of text and transform it into a vector representation with semantic information. By inputting user purchase tendency text data into a context-based encoder model, multiple purchase tendency semantic feature vectors can be obtained. Each feature vector may represent different purchase intentions and preferences, such as product category preference, price sensitivity, and brand preference. These feature vectors can provide a more comprehensive and granular description and analysis of user purchase tendencies. Specifically, the embedding layer of the context-based encoder model containing the embedding layer is used to transform the user purchase tendency text data collected by the web crawler into embedding vectors to obtain a sequence of embedding vectors; the BERT-based converter model of the context-based encoder model containing the embedding layer is used to perform global contextual semantic encoding on the sequence of embedding vectors to obtain multiple purchase tendency semantic feature vectors.

[0032] Furthermore, each feature vector in the multiple semantic feature vectors of purchasing intent represents different aspects or dimensions of purchasing intent information. By performing two-dimensional feature encoding on these feature vectors, they can be mapped into a two-dimensional space, where each dimension represents a feature. This allows for a more intuitive observation of the relationships between different features. Specifically, two-dimensional feature encoding can employ various methods, such as Principal Component Analysis (PCA) or t-SNE (t-Distributed Stochastic Neighbor Embedding). These methods can map high-dimensional feature vectors into a two-dimensional space, preserving the relationships between features.

[0033] Furthermore, each feature vector in multiple purchase intention semantic feature vectors may represent different purchase intention information, such as product category preference, price sensitivity, brand preference, etc. By performing one-dimensional feature encoding on these feature vectors, they can be compressed into a single one-dimensional vector, where each element represents a specific purchase intention phrase. Purchase intention phrase-level semantic feature vectors can more concisely represent users' purchase intentions, making the analysis of purchase intentions more intuitive and interpretable. For example, by comparing the purchase intention phrase-level semantic feature vectors of different users, common purchase intention phrases can be discovered, thereby inferring similarities between users or group behavioral trends. In addition, one-dimensional feature encoding can reduce feature dimensionality, lower computational complexity, and provide simpler input in some machine learning models.

[0034] Specifically, the purchase tendency association semantic feature vector captures the correlation and interaction between different features, reflecting the comprehensive impact of purchase tendency. The purchase tendency phrase-level semantic feature vector, on the other hand, focuses more on capturing information from specific purchase tendency phrases or sentences in the text, providing a finer-grained analysis of purchase tendency. By concatenating these two feature vectors, a richer feature representation can be obtained, encompassing both overall purchase tendency association information and specific purchase tendency phrase information. Specifically, a concatenation function is used to fuse the purchase tendency association semantic feature vector and the purchase tendency phrase-level semantic feature vector to obtain the text understanding feature vector, where the concatenation function is expressed by the formula:

[0035] f(X i ,X j = Relu(W) f [θ(X i ),φ(X j )])

[0036] Among them, W f ,θ(X i ) and φ(X j ) indicates pointwise convolution of the input, ReLU is the activation function, [] indicates concatenation operation, X i X represents the feature value at each position in the semantic feature vector associated with the purchase tendency. j This represents the feature value at each position in the semantic feature vector of the phrase "purchase tendency".

[0037] In a specific embodiment of this application, the two-dimensional feature encoding subunit 1212 for purchasing tendency includes: arranging the plurality of purchasing tendency semantic feature vectors in a two-dimensional manner to obtain a purchasing tendency association feature matrix; and passing the purchasing tendency association feature matrix through a text convolutional neural network to obtain the purchasing tendency association semantic feature vectors. It should be understood that the purpose of arranging the plurality of purchasing tendency semantic feature vectors in a two-dimensional manner to obtain the purchasing tendency association feature matrix is ​​to organize and represent different purchasing tendency features in matrix form, so as to more conveniently perform correlation analysis and pattern recognition between features. The purchasing tendency association feature matrix can arrange multiple purchasing tendency features in a certain order on different rows or columns of the matrix, with each element representing the value of the corresponding feature. By observing and analyzing the patterns and correlations in the matrix, the interrelationships and influences between different purchasing tendency features can be revealed, further understanding the user's purchasing tendency behavior.

[0038] Furthermore, text convolutional neural networks (CNNs) are frequently used in natural language processing for text feature extraction and semantic modeling. By using convolution operations, CNNs can capture local features at different scales and combine and extract these features through pooling operations (such as max pooling). This allows CNNs to effectively learn semantic information and association patterns in text. By inputting a purchase propensity association feature matrix into a CNN, the network can extract association features between purchase propensities through convolution and pooling operations. These association features can capture the semantic relationships and interactions between purchase propensities, thus yielding a purchase propensity association semantic feature vector. This representation better reflects the overall characteristics of purchase propensity and provides richer semantic information. Specifically, each layer of the text convolutional neural network performs the following operations during the forward propagation of the layer: convolution processing is performed on the input data to obtain a text convolutional feature map; mean pooling based on the local feature matrix is ​​performed on the text convolutional feature map to obtain a text pooling feature map; nonlinear activation is performed on the text pooling feature map to obtain a text activation feature map; wherein, the output of the last layer of the text convolutional neural network is the purchase tendency-related semantic feature vector, and the input of the first layer of the text convolutional neural network is the purchase tendency-related feature matrix.

[0039] In a specific embodiment of this application, the one-dimensional feature encoding subunit 1213 for purchasing tendency includes: arranging the multiple semantic feature vectors of purchasing tendency in a one-dimensional manner to obtain a purchasing tendency feature vector; and passing the purchasing tendency feature vector through a multi-scale neighborhood feature extraction module to obtain a phrase-level semantic feature vector of purchasing tendency. It should be understood that by arranging in a one-dimensional manner, multiple semantic feature vectors of purchasing tendency can be connected together in a certain order to form a longer feature vector. This allows the correlation and importance between different features to be taken into account, resulting in a more comprehensive representation of purchasing tendency features. Furthermore, passing the purchasing tendency feature vector through the multi-scale neighborhood feature extraction module to obtain a phrase-level semantic feature vector of purchasing tendency can further extract local correlation patterns and contextual information of the purchasing tendency features to obtain a finer-grained semantic representation. It should be understood that the multi-scale neighborhood feature extraction module typically uses sliding windows or convolutional kernels of different sizes to capture contextual information of different ranges. By extracting neighborhood features at different scales, different levels of semantic correlation can be captured, thereby more comprehensively describing the features of purchasing tendency. By applying a multi-scale neighborhood feature extraction module, the purchase intention feature vector can be decomposed into local feature vectors of different scales, and these feature vectors can capture semantic information of different phrase granularities of purchase intention. Such feature vectors can better represent the details and contextual information of purchase intention, thereby improving the semantic modeling ability of purchase intention. Specifically, the multi-scale neighborhood feature extraction module includes: a first convolutional layer, a second convolutional layer parallel to the first convolutional layer, and a cascaded layer connected to the first convolutional layer and the second convolutional layer, wherein the first convolutional layer uses a one-dimensional convolutional kernel with a first scale, and the second convolutional layer uses a one-dimensional convolutional kernel with a second scale.

[0040] In a specific embodiment of this application, the sales data processing unit 123 includes: normalizing the plurality of Huzhou brush sales feature vectors to obtain a plurality of Huzhou brush sales normalized feature vectors; and concatenating the plurality of Huzhou brush sales normalized feature vectors to obtain the Huzhou brush sales normalized feature vector. It should be understood that different features in the Huzhou brush sales feature vectors may have different value ranges and units, such as sales revenue, sales volume, sales growth rate, etc. Such differences make comparison and analysis between features difficult because their numerical ranges are different and cannot be directly compared. Normalization can map the value ranges of different features to a unified interval, thus eliminating dimensional differences and making different features have similar numerical ranges, facilitating comparison and analysis. Commonly used normalization methods include max-min normalization and standardization. Max-min normalization linearly maps feature values ​​to a specified interval range, while standardization transforms feature values ​​into a distribution with a mean of 0 and a variance of 1 by subtracting the mean and dividing by the standard deviation. Normalizing multiple feature vectors representing the sales situation of Huzhou brushes to obtain multiple normalized feature vectors can eliminate the dimensional differences between different features, making them comparable and interpretable. Specifically, each feature vector representing the sales situation of Huzhou brushes is normalized based on its maximum value to obtain the multiple normalized feature vectors representing the sales situation of Huzhou brushes.

[0041] Furthermore, through cascading, we can connect multiple normalized feature vectors of Huzhou brush sales in a certain order to form a longer feature vector. The advantage of doing this is that it can take into account the correlation and importance between different features, resulting in a more comprehensive normalized feature representation of Huzhou brush sales, thus merging multiple feature vectors into a more comprehensive feature vector to represent the overall Huzhou brush sales situation.

[0042] In the aforementioned intelligent analysis system 100 for commodity sales data, the sales amount prediction module 130 is used to predict whether the sales amount next month will increase or decrease compared to the sales amount this month, based on the text understanding feature vector and the normalized feature vector of the Huzhou brush sales. It should be understood that the text understanding feature vector can contain textual information related to user preferences, such as product descriptions and desired features. The normalized feature vector of the Huzhou brush sales contains multiple numerical features related to sales, such as sales revenue, sales volume, and sales growth rate. These features can reflect the specific situation and trends of sales. Then, the text understanding feature vector and the normalized feature vector of the Huzhou brush sales are fused and optimized. Next, a machine learning model is trained or statistical methods are used to predict whether the sales amount next month will increase or decrease compared to the sales amount this month.

[0043] Figure 4This is a block diagram of the sales revenue prediction module in the intelligent analysis system for commodity sales data according to an embodiment of this application. Figure 4 As shown, in a specific embodiment of this application, the sales amount prediction module 130 includes: a brush pen sales prediction fusion unit 131, used to fuse the text understanding feature vector and the brush pen sales normalized feature vector to obtain a brush pen sales prediction feature matrix; a brush pen sales prediction optimization unit 132, used to optimize the brush pen sales prediction feature matrix based on attention information matching between feature nodes to obtain an optimized brush pen sales prediction feature matrix; and a brush pen sales prediction result generation unit 133, used to pass the optimized brush pen sales prediction feature matrix through a classifier to obtain a classification result, the classification result being used to predict whether the sales amount next month will increase or decrease compared to the sales amount this month. It should be understood that by fusing the text understanding feature vector and the brush pen sales normalized feature vector, text features and numerical features can be combined to form a more comprehensive feature matrix. The advantage of doing so is that it can fully utilize different types of feature information, improving the accuracy and generalization ability of sales prediction.

[0044] Specifically, in the technical solution of this application, in order to fully utilize the correlation and mutual influence between the text understanding feature vector and the brush sales normalized feature vector, and improve the expressive power and predictive performance of the features, it is necessary to pay attention to the correlation between their internal element sub-dimensions in the high-dimensional space when fusing the text understanding feature vector and the brush sales normalized feature vector. Specifically, the text understanding feature vector and the brush sales normalized feature vector often contain different types of information. By focusing on the correlation between their internal element sub-dimensions in the high-dimensional space, the potential connections and shared information between them can be captured. This can enrich the expressive power of the features and provide a more comprehensive feature representation. The text understanding feature vector and the brush sales normalized feature vector may describe the sales situation from different perspectives. By focusing on their correlation, their complementarity can be discovered. Fusing these complementary features can provide a more comprehensive and integrated perspective, helping to better understand and predict the sales situation. Furthermore, when fusing feature vectors, the curse of dimensionality may be encountered, namely, sparsity and redundancy in the high-dimensional feature space. By focusing on the correlations between the sub-dimensions of internal elements, important features relevant to the target task can be identified, while features that are not helpful for prediction can be excluded. This reduces the dimensionality of the feature space, improving the efficiency and generalization performance of the model. Focusing on the correlations between the sub-dimensions of internal elements helps improve the expressive power and predictive performance of features. By fusing the text understanding feature vector and the normalized feature vector of brush sales, and fully utilizing the correlation between them, a more accurate and reliable feature representation can be provided. This can improve the accuracy of brush sales prediction and help to better predict whether the sales amount next month will increase or decrease compared to the sales amount this month. Based on this, in the technical solution of this application, in order to focus on the correlations between the sub-dimensions of internal elements of the text understanding feature vector and the normalized feature vector of brush sales in the high-dimensional space, the brush sales prediction feature matrix is ​​optimized based on attention information matching between feature nodes.

[0045] Specifically, optimizing the Huzhou brush sales prediction feature matrix based on attention information matching between feature nodes to obtain an optimized Huzhou brush sales prediction feature matrix includes: calculating the attention information matching factor between feature nodes between the text understanding feature vector and the Huzhou brush sales normalized feature vector; and weighting the Huzhou brush sales prediction feature matrix with the attention information matching factor between feature nodes to obtain the optimized Huzhou brush sales prediction feature matrix.

[0046] More specifically, the attention information matching factor between feature nodes of the text understanding feature vector and the normalized feature vector of the brush sales is calculated using the following attention information matching factor formula;

[0047] The formula for the attention information matching factor is as follows:

[0048]

[0049] Wherein, V1 represents the text understanding feature vector, and V2 represents the normalized feature vector of the Huzhou brush sales. This represents vector subtraction, v 1i v represents the feature value at the i-th position of the text understanding feature vector. 2i Let $\mathbf{i}$ represent the eigenvalue at the $i$-th position of the normalized eigenvector of the Huzhou brush sales, and $\log$ represent the logarithmic function value to the base 2. F Let λ represent the Frobenius norm of the vector, and let λ represent the attention information matching factor between the feature nodes.

[0050] In other words, to address the aforementioned technical problems, the technical solution of this application uses an information matching method between feature nodes based on an attention mechanism to calculate the attention information matching factor between feature nodes of the text understanding feature vector and the normalized feature vector of the brush sales. The attention information matching factor between feature nodes is used to measure the information between the internal element sub-dimensions of the feature vector in the high-dimensional space, so as to capture the information similarity and difference of different sub-dimensions in the feature vector, thereby constructing a superconvex consistent derivation expression that can reflect the feature manifold structure of the feature vector in the high-dimensional space. This expression can ensure that the projection of the feature vector on different sub-dimensions has convex monotonicity, that is, the manifold distance between feature vectors is proportional to the projection distance on each sub-dimension, thereby enhancing the feature association ability of classification features to feature vectors, that is, classification features can better distinguish feature vectors of different categories, thereby improving the performance of classification tasks.

[0051] Furthermore, a classifier is a machine learning model that can categorize data into different classes based on an input feature vector. In this case, the classifier can use an optimized feature matrix for predicting brush sales as input, classifying increases or decreases in next month's sales amount as different categories based on patterns and regularities in historical sales data. By training the classifier model, it can learn the relationships and patterns between features and classify new feature vectors based on these patterns. The classification results can provide trend information about next month's sales amount relative to this month's sales amount, i.e., predicting whether sales will increase or decrease.

[0052] Figure 5 This is a schematic diagram of the architecture of a smart analysis system for commodity sales data according to an embodiment of this application. Figure 5As shown, the intelligent analysis system 100 for commodity sales data according to an embodiment of this application includes: First, acquiring user purchase tendency text data collected by a web crawler and text data on the sales of Huzhou brushes for a predetermined time period collected from a database, wherein the Huzhou brush sales data includes sales amount, sales quantity, promotional activities, and sales time; then, passing the user purchase tendency text data collected by the web crawler through a context-based encoder model including an embedding layer to obtain multiple purchase tendency semantic feature vectors; next, arranging the multiple purchase tendency semantic feature vectors in a two-dimensional manner to obtain a purchase tendency association feature matrix; then, passing the purchase tendency association feature matrix through a text convolutional neural network to obtain a purchase tendency association semantic feature vector; then, arranging the multiple purchase tendency semantic feature vectors in a one-dimensional manner to obtain a purchase tendency feature vector; further, passing the purchase tendency feature vector through a multi-scale neighborhood feature extraction module to obtain a purchase tendency phrase-level semantic feature vector; and even further, passing the purchase tendency feature vector through a multi-scale neighborhood feature extraction module to obtain a purchase tendency phrase-level semantic feature vector; and finally, passing the purchase tendency text data through a web crawler to obtain a purchase tendency phrase-level semantic feature vector. The text understanding feature vector is obtained by concatenating the semantic feature vector of the tendency association and the semantic feature vector of the purchase tendency phrase granularity. Then, the text data of Huzhou brush sales over a predetermined time period collected from the database is processed through a deep learning-based semantic understanding model to obtain multiple Huzhou brush sales feature vectors. Next, these multiple Huzhou brush sales feature vectors are normalized to obtain multiple Huzhou brush sales normalized feature vectors. Furthermore, these multiple Huzhou brush sales normalized feature vectors are concatenated to obtain a Huzhou brush sales normalized feature vector. Next, the text understanding feature vector and the Huzhou brush sales normalized feature vector are fused to obtain a Huzhou brush sales prediction feature matrix. Next, the Huzhou brush sales prediction feature matrix is ​​optimized based on attention information matching between feature nodes to obtain an optimized Huzhou brush sales prediction feature matrix. Finally, the optimized Huzhou brush sales prediction feature matrix is ​​processed by a classifier to obtain a classification result, which is used to predict whether the sales amount next month will increase or decrease compared to the sales amount this month.

[0053] In summary, this application embodiment first obtains text data on user purchasing tendencies collected by a web crawler and text data on the sales of Huzhou brushes for a predetermined time period collected from a database. The text data on the sales of Huzhou brushes includes sales amount, sales quantity, promotional activities, and sales time. Then, deep learning technology is used to extract features and perform correlation analysis on the two data. Finally, a classifier is used to obtain classification results to predict whether the sales amount in the next month will increase or decrease compared to the sales amount in this month.

[0054] As described above, the intelligent analysis system 100 for commodity sales data according to the embodiments of this application can be implemented in various terminal devices, such as servers deploying intelligent analysis control algorithms for commodity sales data. In one example, the intelligent analysis system 100 for commodity sales data can be integrated into the terminal device as a software module and / or a hardware module. For example, the intelligent analysis system 100 for commodity sales data can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the intelligent analysis system 100 for commodity sales data can also be one of many hardware modules of the terminal device.

[0055] Alternatively, in another example, the intelligent analysis system 100 for product sales data and the terminal device can also be separate devices, and the intelligent analysis system 100 for product sales data can be connected to the terminal device via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.

[0056] Exemplary methods

[0057] Figure 6 This is a flowchart of a product sales data intelligent analysis method according to an embodiment of this application. Figure 6 As shown, the intelligent analysis method for commodity sales data according to an embodiment of this application includes: S110, acquiring user purchase tendency text data collected by a web crawler and Huzhou brush sales data for a predetermined time period collected from a database, wherein the Huzhou brush sales data includes sales amount, sales quantity, promotional activities, and sales time; S120, extracting text understanding feature vectors and Huzhou brush sales normalized feature vectors from the user purchase tendency text data collected by the web crawler and the Huzhou brush sales data for a predetermined time period collected from the database; S130, predicting whether the sales amount next month will increase or decrease compared to the sales amount this month based on the text understanding feature vectors and the Huzhou brush sales normalized feature vectors.

[0058] Here, those skilled in the art will understand that the specific operations of each step in the above-mentioned intelligent analysis method for commodity sales data have been referenced above. Figures 1 to 5 The product sales data intelligent analysis system is described in detail in the description, and therefore, its repeated description will be omitted.

[0059] Exemplary electronic devices

[0060] Based on the above-mentioned intelligent analysis method for commodity sales data, this invention also proposes an electronic device.

[0061] Figure 7 This is a structural block diagram of an electronic device according to an embodiment of the present invention.

[0062] like Figure 7 As shown, the electronic device 10 includes a processor 11 and a memory 13. The processor 11 and the memory 13 are connected, for example, via a bus 12. Optionally, the electronic device 10 may also include a transceiver 14. It should be noted that in practical applications, the transceiver 14 is not limited to one type, and the structure of this electronic device 10 does not constitute a limitation on the embodiments of the present invention.

[0063] Processor 11 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), a FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 11 may also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0064] Bus 12 may include a pathway for transmitting information between the aforementioned components. Bus 12 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 12 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0065] The memory 13 stores a computer program corresponding to the intelligent analysis method for commodity sales data in the above embodiments of the present invention. This computer program is controlled and executed by the processor 11. The processor 11 executes the computer program stored in the memory 13 to implement the content shown in the aforementioned method embodiments.

[0066] Among them, electronic devices 10 include, but are not limited to: mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and vehicle terminals (such as vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 7The electronic device 10 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0067] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0068] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0069] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0070] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0071] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "over," and "on top" of the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0072] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A smart analysis system for commodity sales data, characterized in that, include: The text data acquisition module is used to acquire user purchase tendency text data collected by the crawler and Huzhou brush sales text data collected from the database for a predetermined time period. The Huzhou brush sales text data includes sales amount, sales quantity, promotional activities and sales time. The text data feature extraction module is used to extract text understanding feature vectors and normalized feature vectors of Huzhou brush sales from the user purchase tendency text data collected by the crawler and the Huzhou brush sales data for a predetermined time period collected from the database. The sales amount prediction module is used to predict whether the sales amount in the next month will increase or decrease compared to the sales amount in this month, based on the text understanding feature vector and the normalized feature vector of the Huzhou brush sales.

2. The intelligent analysis system for commodity sales data according to claim 1, characterized in that, The text data feature extraction module includes: The purchase tendency text feature extraction unit is used to extract features from the user purchase tendency text data collected by the crawler to obtain the text understanding feature vector; The sales situation semantic understanding unit is used to process the text data of Huzhou brush sales situation for a predetermined time period collected from the database through a deep learning-based semantic understanding model to obtain the multiple Huzhou brush sales situation feature vectors. The sales data processing unit is used to process the multiple Huzhou brush sales feature vectors to obtain the Huzhou brush sales normalized feature vector.

3. The intelligent analysis system for commodity sales data according to claim 2, characterized in that, The purchase intention text feature extraction unit includes: The purchase tendency semantic understanding subunit is used to process the user purchase tendency text data collected by the crawler through a context-based encoder model containing an embedding layer to obtain multiple purchase tendency semantic feature vectors. A two-dimensional feature encoding subunit for purchasing tendency is used to encode the plurality of semantic feature vectors of purchasing tendency in two dimensions to obtain a semantic feature vector of purchasing tendency association. A one-dimensional feature encoding subunit for purchasing tendency is used to encode the multiple semantic feature vectors of purchasing tendency in one dimension to obtain semantic feature vectors of purchasing tendency at the phrase level. The semantic understanding association subunit is used to concatenate the purchase intention association semantic feature vector and the purchase intention phrase granularity semantic feature vector to obtain the text understanding feature vector.

4. The intelligent analysis system for commodity sales data according to claim 3, characterized in that, The two-dimensional feature encoding subunit for purchasing tendency includes: The multiple semantic feature vectors of purchasing tendency are arranged in two dimensions to obtain the purchasing tendency association feature matrix; The purchase tendency associated feature matrix is ​​processed through a text convolutional neural network to obtain the purchase tendency associated semantic feature vector.

5. The intelligent analysis system for commodity sales data according to claim 4, characterized in that, The one-dimensional feature encoding subunit for purchasing tendency includes: The multiple semantic feature vectors of purchasing tendency are arranged in one dimension to obtain the feature vector of purchasing tendency. The purchase tendency feature vector is processed by a multi-scale neighborhood feature extraction module to obtain the purchase tendency phrase-level semantic feature vector.

6. The intelligent analysis system for commodity sales data according to claim 5, characterized in that, The sales data processing unit includes: The multiple Huzhou brush sales feature vectors are normalized to obtain multiple Huzhou brush sales normalized feature vectors; The multiple normalized feature vectors of Huzhou brush sales are concatenated to obtain the normalized feature vector of Huzhou brush sales.

7. The intelligent analysis system for commodity sales data according to claim 6, characterized in that, The sales revenue prediction module includes: The Huzhou brush sales prediction fusion unit is used to fuse the text understanding feature vector and the Huzhou brush sales normalized feature vector to obtain the Huzhou brush sales prediction feature matrix. The Huzhou brush sales prediction optimization unit is used to optimize the Huzhou brush sales prediction feature matrix based on attention information matching between feature nodes to obtain an optimized Huzhou brush sales prediction feature matrix. The Huzhou brush sales forecast result generation unit is used to pass the optimized Huzhou brush sales forecast feature matrix through a classifier to obtain a classification result, which is used to predict whether the sales amount in the next month will increase or decrease compared to the sales amount in this month.

8. The intelligent analysis system for commodity sales data according to claim 7, characterized in that, The Huzhou brush sales forecasting optimization unit includes: Calculate the attention information matching factor between feature nodes between the text understanding feature vector and the normalized feature vector of the brush sales; The Huzhou brush sales prediction feature matrix is ​​weighted by the attention information matching factor between the feature nodes to obtain an optimized Huzhou brush sales prediction feature matrix.

9. The intelligent analysis system for commodity sales data according to claim 8, characterized in that, The Huzhou brush sales forecasting optimization unit includes: The attention information matching factor between feature nodes of the text understanding feature vector and the normalized feature vector of the brush sales is calculated using the following formula: The formula for the attention information matching factor is as follows: Wherein, V1 represents the text understanding feature vector, and V2 represents the normalized feature vector of the Huzhou brush sales. This represents vector subtraction, v 1i v represents the feature value at the i-th position of the text understanding feature vector. 2i Let $\mathbf{i}$ represent the eigenvalue at the $i$-th position of the normalized eigenvector of the Huzhou brush sales, and $\log$ represent the logarithmic function value to the base 2. F Let λ represent the Frobenius norm of the vector, and let λ represent the attention information matching factor between the feature nodes.

10. A method for intelligent analysis of commodity sales data, characterized in that, include: The system acquires text data on user purchasing preferences collected by a web crawler and text data on the sales of Huzhou brushes for a predetermined time period collected from a database. The text data on the sales of Huzhou brushes includes sales amount, sales quantity, promotional activities, and sales time. Extract text understanding feature vectors and normalized feature vectors of Huzhou brush sales from the user purchase tendency text data collected by the crawler and the Huzhou brush sales data for a predetermined time period collected from the database; Based on the text understanding feature vector and the normalized feature vector of Huzhou brush sales, predict whether the sales amount next month will increase or decrease compared to the sales amount this month.

Images