Business analysis system and method based on multi-dimensional data

Abstract

The invention provides a business analysis system and method based on multi-dimensional data, relates to the field of inventory management, and adopts a data processing and coding mode based on artificial intelligence to perform time sequence feature analysis and association of product sales volume time sequence data and product inventory time sequence data. Therefore, whether the product stock needs to be increased or not is intelligently judged according to the time sequence collaborative interaction representation between the time sequence correlation characteristics of the product sales volume and the time sequence correlation characteristics of the product stock. Therefore, complex time sequence modes and dynamic changes in the time sequence data can be effectively captured, meanwhile, time sequence interaction characteristics between sales data and inventory data are integrated, the inventory level can be adjusted more comprehensively and accurately in real time, and automation of inventory decision making is achieved.

Description

Technical Field

[0001] The present application relates to the field of inventory management, and more specifically, to a business analysis system and method based on multi-dimensional data. Background Art

[0002] With the rapid development of information technology, enterprises and institutions are faced with the generation and accumulation of massive data, which covers all aspects of the enterprise, including but not limited to finance, sales, marketing, customer behavior, supply chain management, etc. In the modern business environment, sales and inventory management are the core links of enterprise operations. Effective sales and inventory management can not only improve the profitability of enterprises, but also enhance customer satisfaction and market competitiveness.

[0003] However, traditional inventory management methods often rely on simple statistical models, which are difficult to cope with the complexity and uncertainty of market changes. Specifically, these simple statistical models are usually static and difficult to capture dynamic changes in time series data. For example, complex patterns such as seasonal changes in inventory, trend changes, and market mutations may not be effectively captured by simple models, resulting in inaccurate forecasts. In addition, these models have difficulty processing multi-dimensional data and complex interactive relationships, and cannot fully consider the diversity and uncertainty of market changes. For example, factors such as promotional activities and holidays may have a significant impact on inventory demand, and simple models cannot effectively integrate this multi-dimensional information, thus affecting the efficiency and accuracy of inventory management.

[0004] Therefore, an optimized business analysis solution based on multi-dimensional data is needed. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a business analysis system and method based on multi-dimensional data, which uses artificial intelligence-based data processing and encoding methods to analyze and associate the time series characteristics of product sales time series data and product inventory time series data, so as to intelligently judge whether it is necessary to increase the product inventory according to the time series collaborative interaction representation between the time series correlation characteristics of product sales and the time series correlation characteristics of product inventory. In this way, the complex time series patterns and dynamic changes in the time series data can be effectively captured, and the time series interaction characteristics between sales and inventory data are integrated, so that the inventory level can be adjusted in real time more comprehensively and accurately, and the automation of inventory decision-making can be realized.

[0006] According to one aspect of the present application, a business analysis system based on multi-dimensional data is provided, which includes: a product sales volume acquisition module, which is used to obtain a time queue of product sales volume from a sales department.

[0007] The product inventory acquisition module is used to obtain the time queue of product inventory from the inventory department.

[0008] The sales volume and inventory time series coding module is used to perform sequence coding on the time series of the product sales volume and the time series of the product inventory to obtain the product sales forward time series correlation characteristics and the product inventory backward time series correlation characteristics.

[0009] The sales volume and inventory feature interaction module is used to perform a time series feature collaborative interaction on the forward time series association feature of the product sales and the backward time series association feature of the product inventory to obtain a deep collaborative interaction representation of the product sales volume and the product inventory time series, wherein the sales volume and inventory feature interaction module includes: a sales volume and inventory feature mapping unit, used to perform feature mapping on the forward time series association feature of the product sales and the backward time series association feature of the product inventory to obtain a modulated product sales forward time series association feature and a modulated product inventory backward time series association feature; a sales volume and inventory compensation interaction unit, used to perform a deviation compensation interaction on the forward time series association feature of the product sales and the backward time series association feature of the product inventory based on the modulated product sales forward time series association feature and the modulated product inventory backward time series association feature to obtain a deep collaborative interaction representation of the product sales volume and the product inventory time series.

[0010] The inventory determination module is used to obtain an optimization result based on the deep collaborative interaction representation of the product sales volume-product inventory time series, and the optimization result is used to indicate whether the product inventory needs to be increased.

[0011] In the above-mentioned business analysis system based on multi-dimensional data, the sales volume and inventory time series encoding module includes: a product sales time series correlation feature extraction unit, which is used to input the time queue of the product sales volume into a product sales volume sequence encoder based on a forward LSTM model to obtain a product sales forward time series correlation feature vector as the product sales forward time series correlation feature; a product inventory time series correlation feature extraction unit, which is used to input the time queue of the product inventory volume into a product inventory sequence encoder based on a backward LSTM model to obtain a product inventory backward time series correlation feature vector as the product inventory backward time series correlation feature.

[0012] In the above-mentioned business analysis system based on multi-dimensional data, the sales volume and inventory feature mapping unit is used to: input the product sales forward time series association feature vector and the product inventory backward time series association feature vector into a co-space mapper based on a fully connected layer to obtain the modulated product sales forward time series association feature vector as the modulated product sales forward time series association feature and the modulated product inventory backward time series association feature vector as the modulated product inventory backward time series association feature, wherein the modulated product sales forward time series association feature vector and the modulated product inventory backward time series association feature vector have the same feature scale.

[0013] In the above-mentioned business analysis system based on multi-dimensional data, the sales volume inventory feature mapping unit is used to: perform matrix-vector multiplication on the weight matrix and the product sales forward time series association feature vector to obtain the product sales forward time series association weight vector; add the product sales forward time series association weight vector and the bias vector by position to obtain the product sales forward time series association coding vector; perform matrix-vector multiplication on the weight matrix and the product inventory backward time series association feature vector to obtain the product inventory backward time series association weight vector; add the product inventory backward time series association weight vector and the bias vector by position to obtain the product inventory backward time series association coding vector; respectively input the product sales forward time series association coding vector and the product inventory backward time series association coding vector into the Relu function for mapping processing to obtain the modulated product sales forward time series association feature vector and the modulated product inventory backward time series association feature vector.

[0014] In the above-mentioned business analysis system based on multi-dimensional data, the sales volume inventory compensation interaction unit includes: a sales volume inventory feature interpolation processing subunit, which is used to interpolate the product sales forward time series association feature vector and the product inventory backward time series association feature vector to obtain the interpolated product sales forward time series association feature vector and the interpolated product inventory backward time series association feature vector, wherein the interpolated product sales forward time series association feature vector has the same characteristic scale as the modulated product sales forward time series association feature vector, and the interpolated product inventory backward time series association feature vector has the same characteristic scale as the modulated product inventory backward time series association feature vector; a sales volume inventory feature mapping deviation measurement subunit, which is used to interpolate the interpolated product sales forward time series association feature vector and the modulated product sales forward time series association feature vector and the interpolated product inventory backward time series association feature vector and The modulated product inventory backward time series association feature vector is input into a feature mapping deviation semantic measurement network to obtain a product sales mapping deviation semantic compensation factor and a product inventory mapping deviation semantic compensation factor; a sales volume inventory feature compensation subunit is used to compensate the modulated product sales forward time series association feature vector and the modulated product inventory backward time series association feature vector based on the product sales mapping deviation semantic compensation factor and the product inventory mapping deviation semantic compensation factor to obtain a compensated product sales forward time series association feature modulation vector and a compensated product inventory backward time series association feature modulation vector; a sales volume inventory feature multidimensional interaction subunit is used to input the compensated product sales forward time series association feature modulation vector and the compensated product inventory backward time series association feature modulation vector into a multidimensional significant interaction network between features to obtain a product sales volume-product inventory time series deep collaborative interaction representation vector as the product sales volume-product inventory time series deep collaborative interaction representation.

[0015] In the above-mentioned business analysis system based on multi-dimensional data, the sales volume inventory feature mapping deviation measurement subunit is used to: calculate the position point division of the modulated product sales forward time series association feature vector and the interpolated product sales forward time series association feature vector to obtain the product sales forward time series offset vector; calculate the logarithmic function value with base two of the absolute value of each eigenvalue of the product sales forward time series offset vector to obtain the product sales forward time series offset logarithmic vector; perform position point multiplication on the modulated product sales forward time series association feature vector and the product sales forward time series offset logarithmic vector, and perform position point multiplication on each eigenvalue in the obtained product sales forward time series offset compensation vector to obtain the product sales forward time series offset compensation value; calculate the natural exponential function value with the natural constant e as the base and the product sales forward time series offset compensation value as the exponential power to obtain the product sales mapping deviation semantic compensation factor.

[0016] In the above-mentioned business analysis system based on multi-dimensional data, the sales volume inventory feature compensation subunit is used to: calculate the sum of the square of the product sales mapping deviation semantic compensation factor and the product sales mapping deviation semantic compensation factor to obtain the total sum of product sales semantic compensation factors; divide the modulated product sales forward time series association feature vector and the total sum of product sales semantic compensation factors by position point to obtain the compensated product sales forward time series association feature modulation vector; calculate the sum of the square of the product inventory mapping deviation semantic compensation factor and the product inventory mapping deviation semantic compensation factor to obtain the total sum of product inventory semantic compensation factors; divide the modulated product inventory backward time series association feature vector and the total sum of product inventory semantic compensation factors by position point to obtain the compensated product inventory backward time series association feature modulation vector.

[0017] In the above-mentioned business analysis system based on multi-dimensional data, the sales volume inventory feature multi-dimensional interaction subunit is used to: respectively calculate the position point subtraction, position point multiplication, position point addition and position point division between the compensated product sales forward time series associated feature modulation vector and the compensated product inventory backward time series associated feature modulation vector to obtain the compensated product sales inventory point subtraction vector, the compensated product sales inventory point multiplication vector, the compensated product sales inventory point addition vector and the compensated product sales inventory point division vector; cascade the compensated product sales inventory point subtraction vector, the compensated product sales inventory point multiplication vector, the compensated product sales inventory point addition vector and the compensated product sales inventory point division vector and input them into a one-dimensional convolutional coding module to obtain a product sales inventory time series multi-dimensional collaborative interaction vector; perform local window-based maximum value pooling processing on the product sales inventory time series multi-dimensional collaborative interaction vector to obtain the product sales volume-product inventory time series deep collaborative interaction representation vector.

[0018] In the above-mentioned business analysis system based on multi-dimensional data, the inventory judgment module is used to: input the product sales-product inventory time series deep collaborative interaction representation vector into the classifier-based business optimization module to obtain the optimization result, and the optimization result is used to indicate whether the product inventory needs to be increased.

[0019] According to another aspect of the present application, a business analysis method based on multi-dimensional data is provided, which includes: obtaining a time queue of product sales from a sales department; obtaining a time queue of product inventory from an inventory department; performing sequence encoding on the time queue of product sales and the time queue of product inventory to obtain a forward time series correlation feature of product sales and a backward time series correlation feature of product inventory; performing time series feature collaborative interaction on the forward time series correlation feature of product sales and the backward time series correlation feature of product inventory to obtain a deep collaborative interaction representation of product sales volume-product inventory time series, including: performing sequence encoding on the forward time series correlation feature of product sales and the backward time series correlation feature of product inventory The feature mapping is performed on the joint feature and the product inventory backward time series association feature to obtain the modulated product sales forward time series association feature and the modulated product inventory backward time series association feature; based on the modulated product sales forward time series association feature and the modulated product inventory backward time series association feature, the product sales forward time series association feature and the product inventory backward time series association feature are subjected to deviation compensation interaction to obtain the product sales volume-product inventory time series deep collaborative interaction representation; based on the product sales volume-product inventory time series deep collaborative interaction representation, an optimization result is obtained, and the optimization result is used to indicate whether the product inventory needs to be increased.

[0020] Compared with the prior art, the business analysis system and method based on multi-dimensional data provided by the present application adopts artificial intelligence-based data processing and coding methods to analyze and associate the time series characteristics of product sales time series data and product inventory time series data, so as to intelligently judge whether it is necessary to increase the product inventory according to the time series collaborative interaction between the time series correlation characteristics of product sales and the time series correlation characteristics of product inventory. In this way, the complex time series patterns and dynamic changes in time series data can be effectively captured, and the time series interaction characteristics between sales and inventory data are integrated, so that the inventory level can be adjusted in real time more comprehensively and accurately, and the automation of inventory decision-making can be realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in 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 described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] In the attached picture: Figure 1 4 is a system block diagram of a business analysis system based on multi-dimensional data according to an embodiment of the present application.

[0023] Figure 2 Schematic diagram of data flow of a business analysis system based on multi-dimensional data according to an embodiment of the present application.

[0024] Figure 3 It is a block diagram of a sales volume and inventory volume time series encoding module in a business analysis system based on multi-dimensional data according to an embodiment of the present application.

[0025] Figure 4 It is a block diagram of the sales volume and inventory feature interaction module in the business analysis system based on multi-dimensional data according to an embodiment of the present application.

[0026] Figure 5 The present invention is a flowchart of a business analysis method based on multi-dimensional data according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0028] With the rapid advancement of information technology, enterprises have accumulated a large amount of data covering multiple fields such as finance, sales, marketing and supply chain. As the core of enterprise operations, sales and inventory management play a decisive role in enhancing the profitability, customer satisfaction and market competitiveness of enterprises. However, traditional inventory management methods are mostly based on static statistical models. These models are insufficient in dealing with market complexity and uncertainty, and it is difficult to accurately capture the time series dynamics of data, such as seasonal fluctuations, trend changes and market mutations. In addition, these models have limited capabilities in processing multi-dimensional data and complex interactive relationships, and fail to fully consider the diversity and uncertainty of market changes, such as the impact of promotional activities and holidays on inventory demand, thus affecting the efficiency and accuracy of inventory management.

[0029] In response to the above technical problems, the technical concept of this application is to obtain the time queue of product sales from the sales department, and the time queue of product inventory from the inventory department, and use artificial intelligence-based data processing and encoding methods to analyze and associate the time series characteristics of the product sales and the product inventory, so as to intelligently judge whether it is necessary to increase the product inventory based on the time series collaborative interaction between the time series correlation characteristics of the product sales and the time series correlation characteristics of the product inventory. In this way, the complex time series patterns and dynamic changes in the time series data can be effectively captured, and the time series interaction characteristics between sales and inventory data are integrated, so that the inventory level can be adjusted in real time more comprehensively and accurately, and the automation of inventory decision-making can be realized.

[0030] Figure 1 4 is a system block diagram of a business analysis system based on multi-dimensional data according to an embodiment of the present application. Figure 2Schematic diagram of data flow of a business analysis system based on multi-dimensional data according to an embodiment of the present application. Figure 1 and Figure 2 As shown, in the business analysis system 100 based on multi-dimensional data, it includes: a product sales volume acquisition module 110, which is used to obtain the time queue of product sales volume from the sales department; a product inventory volume acquisition module 120, which is used to obtain the time queue of product inventory volume from the inventory department; a sales volume inventory volume time series encoding module 130, which is used to respectively perform sequence encoding on the time queue of the product sales volume and the time queue of the product inventory volume to obtain the forward time series correlation feature of product sales and the backward time series correlation feature of product inventory; a sales volume inventory feature interaction module 140, which is used to perform time series feature collaborative interaction on the forward time series correlation feature of product sales and the backward time series correlation feature of product inventory to obtain a product sales volume-product inventory time series deep collaborative interaction representation; an inventory volume judgment module 150, which is used to obtain an optimization result based on the product sales volume-product inventory time series deep collaborative interaction representation, and the optimization result is used to indicate whether the product inventory volume needs to be increased.

[0031] In an embodiment of the present application, the product sales volume acquisition module 110 and the product inventory acquisition module 120 are respectively used to obtain the time queue of product sales volume from the sales department and the time queue of product inventory volume from the inventory department. It should be understood that the time queue data of product sales volume specifically refers to the specific sales value of the product within a period of time (which can be a week, month or quarter), which reflects the market demand for the product. The time queue data of product inventory volume specifically refers to the specific inventory value of the product within a period of time, that is, it indicates the number of available products, which reflects the inventory level of the product. By analyzing the time queue data of product sales volume, it can help predict the future sales trend of the product. When the sales volume of the product continues to rise, and the insufficient inventory level makes it difficult to meet the market demand for the product, it is necessary to increase the product inventory.

[0032] In the embodiment of the present application, the sales volume and inventory time series encoding module 130 is used to perform sequence encoding on the time series of the product sales volume and the time series of the product inventory to obtain the product sales forward time series correlation feature and the product inventory backward time series correlation feature. Specifically, Figure 3 FIG. 1 is a block diagram of a sales volume and inventory volume time series encoding module in a business analysis system based on multi-dimensional data according to an embodiment of the present application. Figure 3As shown, the sales volume and inventory time series encoding module 130 includes: a product sales time series correlation feature extraction unit 131, which is used to input the time queue of the product sales volume into a product sales volume sequence encoder based on a forward LSTM model to obtain a product sales forward time series correlation feature vector as the product sales forward time series correlation feature; a product inventory time series correlation feature extraction unit 132, which is used to input the time queue of the product inventory volume into a product inventory sequence encoder based on a backward LSTM model to obtain a product inventory backward time series correlation feature vector as the product inventory backward time series correlation feature.

[0033] In an embodiment of the present application, the product sales time series correlation feature extraction unit 131 is used to input the time queue of the product sales volume into the product sales volume sequence encoder based on the forward LSTM model to obtain the product sales forward time series correlation feature vector as the product sales forward time series correlation feature. It should be understood that the sales trend usually develops gradually, that is, the future product sales volume is affected by the current sales situation. Considering that the forward LSTM model can learn and memorize the long-term dependencies between different positions in the input data through its unique gating mechanism, and can solve the problem that the ordinary recurrent neural network model is difficult to capture the dynamic changes in long sequence data. Based on this, in order to better and more effectively capture the future trend of product sales volume and the impact of the current state on the future, in the technical solution of the present application, the time queue of the product sales volume is input into the product sales volume sequence encoder based on the forward LSTM model to obtain the product sales forward time series correlation feature vector. That is, the features extracted by the forward LSTM contain time dependencies and patterns from the past to the present, which are helpful in predicting future sales demand.

[0034] In an embodiment of the present application, the product inventory time series correlation feature extraction unit 132 is used to input the time queue of the product inventory into the product inventory sequence encoder based on the backward LSTM model to obtain the product inventory backward time series correlation feature vector as the product inventory backward time series correlation feature. It should be understood that the inventory change usually has historical dependence, that is, the current product inventory is affected by past or historical changes. Therefore, in order to better understand how the current inventory status is affected by past inventory changes and to better understand the trend of inventory time series changes, in the technical solution of the present application, the time queue of the product inventory is input into the product inventory sequence encoder based on the backward LSTM model to capture and extract the product inventory backward time series correlation feature vector. That is, the backward LSTM model processes data from the end of the time series forward, that is, from the most recent inventory data point to the past data point, so that the extracted features contain time dependencies and patterns from the current time point back to the past, and these features help to understand the historical background of inventory changes.

[0035] In the embodiment of the present application, the sales volume inventory feature interaction module 140 is used to perform time series feature collaborative interaction on the product sales forward time series association feature and the product inventory backward time series association feature to obtain a product sales volume-product inventory time series deep collaborative interaction representation. Specifically, Figure 4 FIG. 1 is a block diagram of a sales volume inventory feature interaction module in a business analysis system based on multi-dimensional data according to an embodiment of the present application. Figure 4 As shown, the sales volume inventory feature interaction module 140 includes: a sales volume inventory feature mapping unit 141, which is used to perform feature mapping on the product sales forward time series association feature and the product inventory backward time series association feature to obtain the modulated product sales forward time series association feature and the modulated product inventory backward time series association feature; a sales volume inventory compensation interaction unit 142, which is used to perform deviation compensation interaction on the product sales forward time series association feature and the product inventory backward time series association feature based on the modulated product sales forward time series association feature and the modulated product inventory backward time series association feature to obtain the product sales volume-product inventory time series deep collaborative interaction representation.

[0036] It should be understood that the forward time series correlation feature of product sales and the backward time series correlation feature of product inventory have a certain time series correlation interaction relationship on the time scale. Therefore, in order to capture the deep interaction between sales and inventory, to more comprehensively reflect the dynamic changes between sales and inventory, and to help enterprises more accurately understand the market supply and demand relationship to dynamically adjust inventory levels, in the technical solution of the present application, the forward time series correlation feature of product sales and the backward time series correlation feature of product inventory are subjected to time series feature collaborative interaction to obtain a product sales-product inventory time series deep collaborative interaction representation vector as a product sales-product inventory time series deep collaborative interaction representation.

[0037] Specifically, first, feature mapping is performed on the product sales forward time series association features and the product inventory backward time series association features to obtain modulated product sales forward time series association features and modulated product inventory backward time series association features, so as to standardize and normalize feature vectors of different scales and distributions to ensure that they are compared and processed on the same scale. Specifically, the product sales forward time series association feature vector and the product inventory backward time series association feature vector are input into a co-space mapper based on a fully connected layer to modulate and align the sales time series features and the inventory time series features using the nonlinear mapping relationship of the fully connected layer, and are fused into the same feature space to obtain the modulated product sales forward time series association feature vector as the modulated product sales forward time series association feature and the modulated product inventory backward time series association feature vector as the modulated product inventory backward time series association feature, so that the features that best reflect the essence of the data can be retained while eliminating noise to ensure consistency and relevance in subsequent interactions and analysis.

[0038] Next, considering that some details of the original features may be lost in the process of performing feature space mapping. This lack of information may affect the accuracy of feature interaction analysis, because some key nuances may not be effectively identified. In addition, if there is a certain correlation between the features, independent space transformation may destroy these inherent connections, making it more complicated to identify the interaction between them in subsequent analysis. Therefore, based on the forward time series association features of the modulated product sales and the backward time series association features of the modulated product inventory, the forward time series association features of the product sales and the backward time series association features of the product inventory are subjected to deviation compensation interaction to obtain the deep collaborative interaction representation of the product sales-product inventory time series. Specifically, first, in order to unify data of different time scales to the same scale and ensure that sales and inventory features interact on the same time scale, the forward time series association feature vector of the product sales and the backward time series association feature vector of the product inventory are interpolated so that the scale of the interpolated product sales features and the scale of the modulated product sales features are consistent, and the scale of the interpolated product inventory features and the modulated product inventory features are consistent. Then, the interpolated product sales forward time series association feature vector and the modulated product sales forward time series association feature vector, as well as the interpolated product inventory backward time series association feature vector and the modulated product inventory backward time series association feature vector are input into the feature mapping deviation semantic measurement network to quantify the deviation between the modulated and interpolated feature vectors, determine which regions have experienced excessive or insufficient adjustments, and calculate appropriate compensation factors based on this information to obtain the product sales mapping deviation semantic compensation factor and the product inventory mapping deviation semantic compensation factor. Afterwards, the modulated product sales forward time series association feature vector and the modulated product inventory backward time series association feature vector are compensated based on the product sales mapping deviation semantic compensation factor and the product inventory mapping deviation semantic compensation factor to improve the model's sensitivity to subtle changes in features, and the compensated product sales forward time series association feature modulation vector and the compensated product inventory backward time series association feature modulation vector are obtained. Finally, the compensated features are input into the multi-dimensional significant interaction network between features to more comprehensively reflect the complex dynamic changes and deep interaction between sales and inventory, and the product sales-product inventory time series deep collaborative interaction representation vector is obtained.

[0039] Specifically, in an embodiment of the present application, the sales volume and inventory feature mapping unit is used to: input the product sales forward time series association feature vector and the product inventory backward time series association feature vector into a co-space mapper based on a fully connected layer to obtain the modulated product sales forward time series association feature vector as the modulated product sales forward time series association feature and the modulated product inventory backward time series association feature vector as the modulated product inventory backward time series association feature, wherein the modulated product sales forward time series association feature vector and the modulated product inventory backward time series association feature vector have the same feature scale.

[0040] More specifically, in an embodiment of the present application, the sales volume inventory feature mapping unit is used to: perform matrix-vector multiplication on the weight matrix and the product sales forward time series association feature vector to obtain the product sales forward time series association weight vector; add the product sales forward time series association weight vector and the bias vector by position to obtain the product sales forward time series association coding vector; perform matrix-vector multiplication on the weight matrix and the product inventory backward time series association feature vector to obtain the product inventory backward time series association weight vector; add the product inventory backward time series association weight vector and the bias vector by position to obtain the product inventory backward time series association coding vector; respectively input the product sales forward time series association coding vector and the product inventory backward time series association coding vector into the Relu function for mapping processing to obtain the modulated product sales forward time series association feature vector and the modulated product inventory backward time series association feature vector.

[0041] Specifically, in an embodiment of the present application, the sales volume inventory compensation interaction unit includes: a sales volume inventory feature interpolation processing subunit, which is used to interpolate the product sales forward time series association feature vector and the product inventory backward time series association feature vector to obtain the interpolated product sales forward time series association feature vector and the interpolated product inventory backward time series association feature vector, wherein the interpolated product sales forward time series association feature vector and the modulated product sales forward time series association feature vector have the same feature scale, and the interpolated product inventory backward time series association feature vector and the modulated product inventory backward time series association feature vector have the same feature scale; a sales volume inventory feature mapping deviation measurement subunit, which is used to interpolate the interpolated product sales forward time series association feature vector and the modulated product sales forward time series association feature vector, and the interpolated product inventory backward time series association feature vector and the modulated product inventory backward time series association feature vector The post-modulation product inventory backward time series association feature vector is input into the feature mapping deviation semantic measurement network to obtain the product sales mapping deviation semantic compensation factor and the product inventory mapping deviation semantic compensation factor; the sales volume inventory feature compensation subunit is used to compensate the modulated product sales forward time series association feature vector and the modulated product inventory backward time series association feature vector based on the product sales mapping deviation semantic compensation factor and the product inventory mapping deviation semantic compensation factor to obtain the compensated product sales forward time series association feature modulation vector and the compensated product inventory backward time series association feature modulation vector; the sales volume inventory feature multidimensional interaction subunit is used to input the compensated product sales forward time series association feature modulation vector and the compensated product inventory backward time series association feature modulation vector into the multidimensional significant interaction network between features to obtain the product sales volume-product inventory time series deep collaborative interaction representation vector as the product sales volume-product inventory time series deep collaborative interaction representation.

[0042] More specifically, in an embodiment of the present application, the sales volume inventory feature mapping deviation measurement subunit is used to: calculate the point-by-point division of the modulated product sales forward time series association feature vector and the interpolated product sales forward time series association feature vector to obtain a product sales forward time series offset vector; calculate the logarithmic function value with base two of the absolute value of each eigenvalue of the product sales forward time series offset vector to obtain a product sales forward time series offset logarithmic vector; perform point-by-point multiplication of the modulated product sales forward time series association feature vector and the product sales forward time series offset logarithmic vector, and perform point-by-point multiplication of each eigenvalue in the obtained product sales forward time series offset compensation vector to obtain a product sales forward time series offset compensation value; calculate a natural exponential function value with the natural constant e as the base and the product sales forward time series offset compensation value as an exponential power to obtain the product sales mapping deviation semantic compensation factor.

[0043] More specifically, in an embodiment of the present application, the sales volume inventory feature compensation subunit is used to: calculate the sum of the square of the product sales mapping deviation semantic compensation factor and the product sales mapping deviation semantic compensation factor to obtain the total sum of product sales semantic compensation factors; divide the modulated product sales forward timing association feature vector and the total sum of the product sales semantic compensation factors by position point to obtain the compensated product sales forward timing association feature modulation vector; calculate the sum of the square of the product inventory mapping deviation semantic compensation factor and the product inventory mapping deviation semantic compensation factor to obtain the total sum of product inventory semantic compensation factors; divide the modulated product inventory backward timing association feature vector and the total sum of the product inventory semantic compensation factors by position point to obtain the compensated product inventory backward timing association feature modulation vector.

[0044] More specifically, in an embodiment of the present application, the sales volume inventory feature multidimensional interaction subunit is used to: respectively calculate the position point subtraction, position point multiplication, position point addition and position point division between the compensated product sales forward time series associated feature modulation vector and the compensated product inventory backward time series associated feature modulation vector to obtain the compensated product sales inventory point subtraction vector, the compensated product sales inventory point multiplication vector, the compensated product sales inventory point addition vector and the compensated product sales inventory point division vector; cascade the compensated product sales inventory point subtraction vector, the compensated product sales inventory point multiplication vector, the compensated product sales inventory point addition vector and the compensated product sales inventory point division vector, and input them into a one-dimensional convolutional coding module to obtain a product sales inventory time series multidimensional collaborative interaction vector; perform local window-based maximum value pooling processing on the product sales inventory time series multidimensional collaborative interaction vector to obtain the product sales volume-product inventory time series deep collaborative interaction representation vector.

[0045] In the embodiment of the present application, specifically, the sales volume inventory feature interaction module is used to: perform time series feature collaborative interaction on the product sales forward time series association feature and the product inventory backward time series association feature according to the following formula; wherein the formula is: ;in, and are respectively the product sales forward time series correlation feature vector and the product inventory backward time series correlation feature vector, and are the weight matrix and bias vector respectively, yes function, and They are the forward time series correlation feature vector of the modulated product sales and the backward time series correlation feature vector of the modulated product inventory, To interpolate the eigenvectors, and are the forward time series correlation feature vector of product sales after interpolation and the backward time series correlation feature vector of product inventory after interpolation, For the The eigenvalues ​​at each position in , For the The eigenvalues ​​at each position in , For the The eigenvalues ​​at each position in , For the The eigenvalues ​​at each position in , It means to calculate the logarithmic function value with base 2. is the and stated The number of eigenvalues ​​in , It means calculating the exponential function value with the natural constant e as the base. is the semantic compensation factor for product sales mapping deviation, is the semantic compensation factor for product inventory mapping deviation, is the compensation product sales forward time series correlation characteristic modulation vector, is the backward temporal correlation feature modulation vector of product inventory after compensation, , , and They are point-by-point subtraction, point-by-point multiplication, point-by-point addition, and point-by-point division. For cascade processing, is a one-dimensional convolutional coding operation, is the maximum pooling operation, It is the representation vector of the deep collaborative interaction of the product sales volume and product inventory time series.

[0046] In an embodiment of the present application, the inventory judgment module 150 is used to obtain an optimization result based on the deep collaborative interaction representation of the product sales volume-product inventory time series, and the optimization result is used to indicate whether the product inventory needs to be increased. Specifically, in an embodiment of the present application, the inventory judgment module is used to: input the product sales volume-product inventory time series deep collaborative interaction representation vector into the classifier-based business optimization module to obtain the optimization result, and the optimization result is used to indicate whether the product inventory needs to be increased. That is, the product sales volume-product inventory time series deep collaborative interaction representation obtained by the collaborative interaction of the product sales forward time series association feature vector and the product inventory backward time series association feature vector is used for classification processing to intelligently determine whether the product inventory needs to be increased. In this way, the complex time series patterns and dynamic changes in the time series data can be effectively captured, and the time series interaction characteristics between the sales and inventory data are integrated, so that the inventory level can be adjusted in real time more comprehensively and accurately, and the automation of inventory decision-making can be realized. In particular, in a specific embodiment of the present application, the product sales-product inventory time series deep collaborative interaction representation vector is input into a classifier-based business optimization module to obtain the optimization result, and the optimization result is used to indicate whether the product inventory needs to be increased, including: using the fully connected layer of the classifier to perform fully connected encoding on the product sales-product inventory time series deep collaborative interaction representation vector to obtain a fully connected encoded feature vector of the product sales-product inventory time series deep collaborative interaction representation; inputting the fully connected encoded feature vector of the product sales-product inventory time series deep collaborative interaction representation into the Softmax classification function of the classifier to obtain the probability value of the product sales-product inventory time series deep collaborative interaction representation vector belonging to each classification label, the classification labels including those used to indicate that the product inventory currently needs to be increased and those used to indicate that the product inventory currently does not need to be increased; the classification label corresponding to the largest of the probability values ​​is determined as the optimization result.

[0047] In particular, when the product sales forward time series correlation feature vector and the product inventory backward time series correlation feature vector represent the short-range-long-range bidirectional time series correlation characteristics of product sales and product inventory respectively, when performing feature interaction based on feature mapping deviation compensation, the product sales-product inventory time series deep collaborative interaction representation vector will also have a compensatory difference in the time series correlation feature mapping deviation based on the time series distribution difference of heterogeneous data sources, thereby causing the diversity of interactive feature distribution of deviation compensation. As a result, when the product sales-product inventory time series deep collaborative interaction representation vector is subjected to classification probability regression through a classifier-based business optimization module, the accuracy of the classification result will be affected.

[0048] Preferably, inputting the product sales-product inventory time series deep collaborative interaction representation vector into a classifier-based business optimization module to obtain an optimization result includes: determining a feature mean and a feature variance of the product sales-product inventory time series deep collaborative interaction representation vector, and dividing the feature mean by the feature variance to obtain a product sales-product inventory time series deep collaborative interaction probability statistic value: wherein the process can be expressed as: ;in, is the feature mean of the deep collaborative interaction representation vector of the product sales-product inventory time series, is the characteristic variance of the deep collaborative interaction representation vector of the product sales volume-product inventory time series, It is the probability statistic value of the deep collaborative interaction of the product sales volume-product inventory time series.

[0049] The product sales-product inventory time series deep collaborative interaction representation vector is point-multiplied with the inverse of the maximum eigenvalue in the product sales-product inventory time series deep collaborative interaction representation vector to obtain the product sales-product inventory time series deep collaborative interaction probability constraint vector: wherein the process can be expressed as: ;in, is the product sales-product inventory time series deep collaborative interaction representation vector, The reciprocal of the maximum eigenvalue in the deep collaborative interaction representation vector of the product sales-product inventory time series, It means point multiplication by position. is the probability constraint vector of the deep collaborative interaction of the product sales volume-product inventory time series.

[0050] After dot-wise addition of the product sales-product inventory time series deep collaborative interaction probability constraint vector and the product sales-product inventory time series deep collaborative interaction probability statistical value, the logarithmic value with base 2 is calculated to obtain the product sales-product inventory time series deep collaborative interaction information interaction vector: wherein, the process can be expressed as: ;in, is the product sales-product inventory time series deep collaborative interaction probability constraint vector, is the probability statistic value of the deep collaborative interaction of the product sales volume and product inventory time series, Indicates adding by position point. represents the logarithmic function with base 2, It is the product sales volume-product inventory time series deep collaborative interaction information interaction vector.

[0051] After performing point subtraction on the product sales-product inventory time series deep collaborative interaction probability constraint vector using the product sales-product inventory time series deep collaborative interaction probability statistics value, the inverse of each eigenvalue is calculated to obtain the product sales-product inventory time series deep collaborative interaction sequence constraint vector: ;in, is the product sales-product inventory time series deep collaborative interaction probability constraint vector, is the probability statistic value of the deep collaborative interaction of the product sales volume and product inventory time series, To calculate the inverse of each eigenvalue in a vector, It means to subtract by position point. is the deep collaborative interaction sequence constraint vector of the product sales volume-product inventory time series.

[0052] The product sales volume-product inventory timing deep collaborative interaction information interaction vector and the product sales volume-product inventory timing deep collaborative interaction sequence constraint vector are interpolated to obtain an optimized product sales volume-product inventory timing deep collaborative interaction representation vector; the optimized product sales volume-product inventory timing deep collaborative interaction representation vector is input into the classifier-based business optimization module to obtain the optimization result.

[0053] That is, considering that the weight matrix during classification probability regression acts on the product sales-product inventory time series deep collaborative interaction representation vector, the distribution diversity of the product sales-product inventory time series deep collaborative interaction representation vector causes the uncertainty of the weight stimulus parameters of the weight matrix, resulting in the lack of posterior probability density inference of the product sales-product inventory time series deep collaborative interaction representation vector through the weight matrix, which will affect the accuracy of the probability regression results.

[0054] Therefore, the probability statistical characteristics of the product sales-product inventory time series deep collaborative interaction representation vector are used to simulate the mesoscale interaction structure under probability constraints between the eigenvalue scale and the eigenvector scale of the product sales-product inventory time series deep collaborative interaction representation vector, so as to construct a bidirectional latent variable motif based on the mesoscale short sequence relative to the probability statistical value to perform mesoscale bidirectional migration, and perform posterior recovery based on the sequence constraints on the interactive information, thereby improving the convergence effect in the probability density domain and improving the accuracy of the optimization results obtained by inputting the product sales-product inventory time series deep collaborative interaction representation vector into the classifier-based business optimization module. In this way, the complex time series patterns and dynamic changes in time series data can be effectively captured, while the time series interaction characteristics between sales and inventory data are integrated, so that the inventory level can be adjusted in real time more comprehensively and accurately, and the automation of inventory decision-making can be realized.

[0055] In summary, the multi-dimensional data-based business analysis system 100 according to the embodiment of the present application is explained, which uses artificial intelligence-based data processing and coding methods to analyze and associate the time series characteristics of product sales time series data and product inventory time series data, so as to intelligently determine whether it is necessary to increase the product inventory according to the time series collaborative interaction between the time series correlation characteristics of product sales and the time series correlation characteristics of product inventory. In this way, the complex time series patterns and dynamic changes in the time series data can be effectively captured, and the time series interaction characteristics between sales and inventory data are integrated, so that the inventory level can be adjusted in real time more comprehensively and accurately, and the automation of inventory decision-making can be realized.

[0056] Figure 5 FIG. 1 is a flow chart of a business analysis method based on multi-dimensional data according to an embodiment of the present application. Figure 5 As shown, in the business analysis method based on multi-dimensional data, it includes: S110, obtaining the time queue of product sales from the sales department; S120, obtaining the time queue of product inventory from the inventory department; S130, respectively performing sequence encoding on the time queue of product sales and the time queue of product inventory to obtain the forward time series correlation feature of product sales and the backward time series correlation feature of product inventory; S140, performing time series feature collaborative interaction on the forward time series correlation feature of product sales and the backward time series correlation feature of product inventory to obtain the deep collaborative interaction representation of product sales-product inventory time series; S150, based on the product sales-product inventory time series, A deep collaborative interaction representation of the sequence is obtained to obtain an optimization result, and the optimization result is used to indicate whether it is necessary to increase the product inventory. Step S140 includes: performing feature mapping on the product sales forward time series association characteristics and the product inventory backward time series association characteristics to obtain the modulated product sales forward time series association characteristics and the modulated product inventory backward time series association characteristics; based on the modulated product sales forward time series association characteristics and the modulated product inventory backward time series association characteristics, performing deviation compensation interaction on the product sales forward time series association characteristics and the product inventory backward time series association characteristics to obtain the product sales-product inventory time series deep collaborative interaction representation.

[0057] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned business analysis method based on multi-dimensional data have been referred to above. Figures 1 to 4 The description of the business analysis system based on multi-dimensional data has been introduced in detail, and therefore, its repeated description will be omitted.

[0058] In summary, the business analysis method based on multi-dimensional data according to the embodiment of the present application is explained, which uses artificial intelligence-based data processing and coding to analyze and associate the time series characteristics of product sales time series data and product inventory time series data, so as to intelligently judge whether it is necessary to increase the product inventory according to the time series collaborative interaction between the time series correlation characteristics of product sales and the time series correlation characteristics of product inventory. In this way, the complex time series patterns and dynamic changes in the time series data can be effectively captured, and the time series interaction characteristics between sales and inventory data are integrated, so that the inventory level can be adjusted in real time more comprehensively and accurately, and the automation of inventory decision-making can be realized.

[0059] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details of the above application are only for the purpose of illustration and ease of understanding, rather than limitation, and the above details do not limit the present application to the use of the above specific details to be implemented.

Claims

1. A business analysis system based on multi-dimensional data, characterized in that: include: Product sales volume acquisition module, used to obtain the time queue of product sales volume from the sales department; A product inventory acquisition module is used to obtain a time queue of product inventory from the inventory department; a sales inventory time series encoding module is used to perform sequence encoding on the time queue of product sales and the time queue of product inventory respectively to obtain product sales forward time series correlation features and product inventory backward time series correlation features; a sales inventory feature interaction module is used to perform time series feature collaborative interaction on the product sales forward time series correlation features and the product inventory backward time series correlation features to obtain a product sales-product inventory time series deep collaborative interaction representation, wherein the sales inventory feature interaction module includes: a sales inventory feature mapping unit, which is used to perform time series feature collaborative interaction on the product sales forward time series correlation features and the product inventory backward time series correlation features The product inventory backward time series association features are feature mapped to obtain the modulated product sales forward time series association features and the modulated product inventory backward time series association features; a sales volume inventory compensation interaction unit is used to perform deviation compensation interaction on the product sales forward time series association features and the product inventory backward time series association features based on the modulated product sales forward time series association features and the modulated product inventory backward time series association features to obtain the product sales volume-product inventory time series deep collaborative interaction representation; an inventory judgment module is used to obtain an optimization result based on the product sales volume-product inventory time series deep collaborative interaction representation, and the optimization result is used to indicate whether the product inventory needs to be increased.

2. The business analysis system based on multi-dimensional data according to claim 1, characterized in that: The sales volume and inventory time series encoding module includes: a product sales time series correlation feature extraction unit, which is used to input the time queue of the product sales volume into a product sales volume sequence encoder based on a forward LSTM model to obtain a product sales forward time series correlation feature vector as the product sales forward time series correlation feature; a product inventory time series correlation feature extraction unit, which is used to input the time queue of the product inventory volume into a product inventory sequence encoder based on a backward LSTM model to obtain a product inventory backward time series correlation feature vector as the product inventory backward time series correlation feature.

3. The business analysis system based on multi-dimensional data according to claim 2, characterized in that: The sales volume and inventory feature mapping unit is used to: input the product sales forward time series association feature vector and the product inventory backward time series association feature vector into a co-space mapper based on a fully connected layer to obtain the modulated product sales forward time series association feature vector as the modulated product sales forward time series association feature and the modulated product inventory backward time series association feature vector as the modulated product inventory backward time series association feature, wherein the modulated product sales forward time series association feature vector and the modulated product inventory backward time series association feature vector have the same feature scale.

4. The business analysis system based on multi-dimensional data according to claim 3, characterized in that: The sales volume inventory feature mapping unit is used to: perform matrix-vector multiplication on the weight matrix and the product sales forward time series association feature vector to obtain the product sales forward time series association weight vector; add the product sales forward time series association weight vector and the bias vector by position to obtain the product sales forward time series association coding vector; perform matrix-vector multiplication on the weight matrix and the product inventory backward time series association feature vector to obtain the product inventory backward time series association weight vector; add the product inventory backward time series association weight vector and the bias vector by position to obtain the product inventory backward time series association coding vector; The product sales forward time series association coding vector and the product inventory backward time series association coding vector are respectively input into the Relu function for mapping processing to obtain the modulated product sales forward time series association feature vector and the modulated product inventory backward time series association feature vector.

5. The business analysis system based on multi-dimensional data according to claim 4, characterized in that: The sales volume inventory compensation interaction unit includes: a sales volume inventory feature interpolation processing subunit, which is used to interpolate the product sales forward time series association feature vector and the product inventory backward time series association feature vector to obtain the interpolated product sales forward time series association feature vector and the interpolated product inventory backward time series association feature vector, wherein the interpolated product sales forward time series association feature vector and the modulated product sales forward time series association feature vector have the same feature scale, and the interpolated product inventory backward time series association feature vector and the modulated product inventory backward time series association feature vector have the same feature scale; a sales volume inventory feature mapping deviation measurement subunit, which is used to interpolate the interpolated product sales forward time series association feature vector and the modulated product sales forward time series association feature vector, and the interpolated product inventory backward time series association feature vector and the modulated product inventory backward time series association feature vector. The backward time series association feature vector is input into the feature mapping deviation semantic measurement network to obtain the product sales mapping deviation semantic compensation factor and the product inventory mapping deviation semantic compensation factor; the sales volume inventory feature compensation subunit is used to compensate the modulated product sales forward time series association feature vector and the modulated product inventory backward time series association feature vector based on the product sales mapping deviation semantic compensation factor and the product inventory mapping deviation semantic compensation factor to obtain the compensated product sales forward time series association feature modulation vector and the compensated product inventory backward time series association feature modulation vector; the sales volume inventory feature multidimensional interaction subunit is used to input the compensated product sales forward time series association feature modulation vector and the compensated product inventory backward time series association feature modulation vector into the multidimensional significant interaction network between features to obtain the product sales volume-product inventory time series deep collaborative interaction representation vector as the product sales volume-product inventory time series deep collaborative interaction representation.

6. The business analysis system based on multi-dimensional data according to claim 5, characterized in that: The sales volume inventory feature mapping deviation measurement subunit is used to: calculate the position point division of the modulated product sales forward time series association feature vector and the interpolated product sales forward time series association feature vector to obtain a product sales forward time series offset vector; calculate the logarithmic function value with base two of the absolute value of each eigenvalue of the product sales forward time series offset vector to obtain a product sales forward time series offset logarithmic vector; perform position point multiplication on the modulated product sales forward time series association feature vector and the product sales forward time series offset logarithmic vector, and perform position point multiplication on each eigenvalue in the obtained product sales forward time series offset compensation vector to obtain a product sales forward time series offset compensation value; calculate the natural exponential function value with the natural constant e as the base and the product sales forward time series offset compensation value as the exponential power to obtain the product sales mapping deviation semantic compensation factor.

7. The business analysis system based on multi-dimensional data according to claim 6, characterized in that: The sales volume inventory feature compensation subunit is used to: calculate the sum of the square of the product sales mapping deviation semantic compensation factor and the product sales mapping deviation semantic compensation factor to obtain the total sum of the product sales semantic compensation factors; divide the modulated product sales forward time series association feature vector and the total sum of the product sales semantic compensation factors by position points to obtain the compensated product sales forward time series association feature modulation vector; Calculating the sum of the square of the product inventory mapping deviation semantic compensation factor and the product inventory mapping deviation semantic compensation factor to obtain a total sum of product inventory semantic compensation factors; The modulated product inventory backward temporal sequence association feature vector and the total sum of the product inventory semantic compensation factor are divided by position points to obtain the compensated product inventory backward temporal sequence association feature modulation vector.

8. The business analysis system based on multi-dimensional data according to claim 7, characterized in that: The sales volume inventory feature multidimensional interaction subunit is used to respectively calculate the position point subtraction, position point multiplication, position point addition and position point division between the compensated product sales forward time series correlation feature modulation vector and the compensated product inventory backward time series correlation feature modulation vector to obtain the compensated product sales inventory point subtraction vector, the compensated product sales inventory point multiplication vector, the compensated product sales inventory point addition vector and the compensated product sales inventory point division vector; The compensated product sales inventory point subtraction vector, the compensated product sales inventory point multiplication vector, the compensated product sales inventory point addition vector and the compensated product sales inventory point division vector are cascaded and input into a one-dimensional convolutional coding module to obtain a product sales inventory time series multi-dimensional collaborative interaction vector; the product sales inventory time series multi-dimensional collaborative interaction vector is subjected to local window-based maximum value pooling processing to obtain the product sales volume-product inventory time series deep collaborative interaction representation vector.

9. The business analysis system based on multi-dimensional data according to claim 8, characterized in that: The inventory judgment module is used to: input the product sales-product inventory time series deep collaborative interaction representation vector into the classifier-based business optimization module to obtain the optimization result, and the optimization result is used to indicate whether the product inventory needs to be increased.

10. A business analysis method based on multi-dimensional data, characterized in that: include: Get the time queue of product sales from the sales department; Get the time queue of product inventory from the inventory department; Sequentially encoding the time queue of the product sales volume and the time queue of the product inventory volume to obtain a forward time series correlation feature of the product sales and a backward time series correlation feature of the product inventory; The forward time series correlation feature of the product sales and the backward time series correlation feature of the product inventory are subjected to time series feature collaborative interaction to obtain a product sales volume-product inventory time series deep collaborative interaction representation, including: feature mapping the forward time series correlation feature of the product sales and the backward time series correlation feature of the product inventory to obtain a modulated product sales forward time series correlation feature and a modulated product inventory backward time series correlation feature; based on the modulated product sales forward time series correlation feature and the modulated product inventory backward time series correlation feature, the forward time series correlation feature of the product sales and the backward time series correlation feature of the product inventory are subjected to deviation compensation interaction to obtain a product sales volume-product inventory time series deep collaborative interaction representation; based on the product sales volume-product inventory time series deep collaborative interaction representation, an optimization result is obtained, and the optimization result is used to indicate whether the product inventory needs to be increased.

Images