Business opportunity prediction method, electronic device and computer program product

Abstract

The invention provides a business opportunity prediction method, electronic equipment and a computer program product. The business opportunity prediction method comprises the steps that time series data are acquired, and the time series data are used for representing various target display data of an application program in each date of a plurality of dates and various target user operation data for display content of the application program; performing feature extraction on the time sequence data to obtain a first time sequence feature; adding the first time sequence feature and the time sequence data to obtain first data; and performing business opportunity prediction based on the first data to obtain a business opportunity prediction result.

Description

Technical Field

[0001] The present disclosure relates to technical fields such as artificial intelligence, and more particularly to a business opportunity prediction method, electronic device, and computer program product. Background Art

[0002] In the digital business environment, various applications have become important platforms for enterprises to interact with users. Enterprises urgently need to accurately predict the volume of business opportunities in order to rationally plan resources and formulate marketing strategies.

[0003] In related technologies, business opportunity volume is predicted using empirical formulas. However, since the fluctuation of business opportunity volume is affected by multiple nonlinear factors rather than a simple linear relationship, the accuracy of the business opportunity volume predicted using empirical formulas is low. Summary of the Invention

[0004] The present disclosure provides a business opportunity volume prediction method, electronic device, storage medium, and program product.

[0005] According to one aspect of the present disclosure, a method for predicting business opportunity volume is provided, comprising: acquiring time series data, wherein the time series data is used to characterize a variety of target display data of an application and a variety of target user operation data for the display content of the application on each of a plurality of dates; performing feature extraction on the time series data to obtain a first time series feature; adding the first time series feature to the time series data to obtain first data; and performing business opportunity volume prediction based on the first data to obtain a business opportunity volume prediction result.

[0006] According to the business opportunity volume prediction method of at least one embodiment of the present disclosure, feature extraction is performed on the time series data to obtain a first time series feature, including: performing a first dilated causal convolution on the time series data to obtain a first convolution result; performing a first weight normalization on the first convolution result to obtain a first normalized result; performing a first nonlinear processing on the first normalized result to obtain a first processing result; and obtaining the first time series feature based on the first processing result.

[0007] According to the business opportunity volume prediction method of at least one embodiment of the present disclosure, the first time series feature is obtained based on the first processing result, including: performing a second dilated causal convolution on the first processing result to obtain a second convolution result; performing a second weight normalization on the second convolution result to obtain a second normalized result; performing a second nonlinear processing on the second normalized result to obtain a second processing result; and obtaining the first time series feature based on the second processing result.

[0008] According to the business opportunity quantity prediction method of at least one embodiment of the present disclosure, business opportunity quantity prediction is performed based on the first data to obtain a business opportunity quantity prediction result, including: encoding the first data based on a self-attention mechanism to obtain an encoding feature; decoding the encoding feature based on the self-attention mechanism to obtain a decoding feature; determining a target feature based on the decoding feature; and performing business opportunity quantity prediction based on the target feature to obtain the business opportunity quantity prediction result.

[0009] According to at least one embodiment of the present disclosure, a business opportunity prediction method determines a target feature based on the decoded feature, including: capturing the temporal dependency in the decoded feature through a long short-term memory network, and using the captured second temporal feature as the target feature.

[0010] According to the business opportunity prediction method of at least one embodiment of the present disclosure, the target feature is determined based on the decoded feature, including: extracting a first channel attention weight from the decoded feature; performing weighted calculation on the decoded feature based on the first channel attention weight to obtain a first channel attention feature; extracting a first spatial attention weight from the first channel attention feature; and performing weighted calculation on the first channel attention feature based on the first spatial attention weight, and using the calculation result as the target feature.

[0011] According to the business opportunity prediction method of at least one embodiment of the present disclosure, the first channel attention weight is extracted from the decoding feature, including: performing global average pooling and global maximum pooling on the decoding feature respectively; processing the global average pooling results and the global maximum pooling results respectively through a shared multi-layer perceptron; fusing the processed global average pooling result and the processed global maximum pooling result to obtain a fused feature; and mapping the fused feature through a first activation function to obtain the first channel attention weight.

[0012] According to the business opportunity prediction method of at least one embodiment of the present disclosure, a first spatial attention weight is extracted from the first channel attention feature, including: performing global average pooling and global maximum pooling on the first channel attention feature along the feature point channel direction; splicing the global average pooling result and the global maximum pooling result of the same channel to obtain a spliced ​​feature; convolving the spliced ​​feature; and mapping the convolved spliced ​​feature through a second activation function to obtain the first spatial attention weight.

[0013] According to at least one embodiment of the present disclosure, a business opportunity prediction method determines a target feature based on the decoding feature, including: capturing the temporal dependency in the decoding feature through a long short-term memory network to obtain a second temporal feature; extracting a second channel attention weight from the second temporal feature; performing weighted calculation on the second temporal feature based on the second channel attention weight to obtain a second channel attention feature; extracting a second spatial attention weight from the second channel attention feature; and performing weighted calculation on the second channel attention feature based on the second spatial attention weight, and using the calculation result as the target feature.

[0014] According to the business opportunity volume prediction method of at least one embodiment of the present disclosure, before obtaining the time series data, it also includes: obtaining different types of target data stored on each of a plurality of dates, wherein the different types of target data include different types of target display data of the application and different types of target user operation data for the display content of the application, and the storage period of the target data is less than or equal to one day; for each date of the plurality of dates, for each type of target data, summing up the same type of target data stored multiple times on the date; and generating the time series data based on the order of the dates according to the result of summing up each type of target data on each date of the plurality of dates.

[0015] According to the business opportunity volume prediction method of at least one embodiment of the present disclosure, before obtaining the different types of target data stored on each of a plurality of dates, it also includes: obtaining the different types of data and business opportunity volume stored on each of a plurality of dates, wherein the different types of data include different types of display data of the application and different types of user operation data for the display content of the application; calculating the correlation coefficient between each type of data and the business opportunity volume based on the different types of data and business opportunity volume stored on each of the plurality of dates; and taking the data with a correlation coefficient greater than or equal to a correlation coefficient threshold as the target data.

[0016] According to another aspect of the present disclosure, an electronic device is provided, comprising: a memory storing execution instructions; and a processor executing the execution instructions stored in the memory, so that the processor executes the business opportunity volume prediction method of any embodiment of the present disclosure.

[0017] According to another aspect of the present disclosure, a readable storage medium is provided, wherein the readable storage medium stores execution instructions, and when the execution instructions are executed by a processor, the execution instructions are used to implement the business opportunity volume prediction method of any embodiment of the present disclosure.

[0018] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the method for predicting business opportunity volume according to any embodiment of the present disclosure is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings illustrate exemplary embodiments of the present disclosure and together with the description serve to explain the principles of the present disclosure. These drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification.

[0020] Figure 1 It is a flowchart of a business opportunity volume prediction method according to an embodiment of the present disclosure.

[0021] Figure 2 FIG. 4 is a schematic diagram of a process for extracting first time series features according to an embodiment of the present disclosure.

[0022] Figure 3 FIG. 4 is a visualization example diagram of a dilated causal convolutional layer according to an embodiment of the present disclosure.

[0023] Figure 4 FIG. 4 is a schematic diagram of a process for extracting first time series features according to another embodiment of the present disclosure.

[0024] Figure 5 FIG. 4 is a schematic diagram of a process for predicting business opportunity volume according to an embodiment of the present disclosure.

[0025] Figure 6 FIG. 4 is a schematic diagram of a process for determining target features according to an embodiment of the present disclosure.

[0026] Figure 7 2 is a schematic diagram of a process for determining the first channel attention weight according to an embodiment of the present disclosure.

[0027] Figure 8 It is a visual diagram of the first channel attention weight determination process of an embodiment of the present disclosure.

[0028] Figure 9 This is a schematic diagram of the process of determining the first spatial attention weight according to an embodiment of the present disclosure.

[0029] Figure 10 It is a visual diagram of the first spatial attention weight determination process of an embodiment of the present disclosure.

[0030] Figure 11 FIG. 4 is a schematic diagram of a process for determining target features according to another embodiment of the present disclosure.

[0031] Figure 12 FIG. 4 is a schematic diagram of a process for generating time series data according to an embodiment of the present disclosure.

[0032] Figure 13 FIG. 4 is a schematic diagram of a process for determining target data according to an embodiment of the present disclosure.

[0033] Figure 14 It is a structural diagram of a business opportunity volume prediction model according to an embodiment of the present disclosure.

[0034] Figure 15 Schematic diagram of the structure of a TCN module according to an embodiment of the present disclosure.

[0035] Figure 16 Schematic diagram of the structure of a CBAM module according to an embodiment of the present disclosure.

[0036] Figure 17 It is a flowchart of a business opportunity quantity prediction method according to another embodiment of the present disclosure.

[0037] Figure 18 It is a schematic block diagram of the structure of a business opportunity volume prediction device according to an embodiment of the present disclosure.

[0038] Figure 19 It is a schematic block diagram of the structure of an electronic device according to one embodiment of the present disclosure. DETAILED DESCRIPTION

[0039] The present disclosure is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are intended only to illustrate the relevant content and are not intended to limit the present disclosure. It should also be noted that, for ease of description, only the portions relevant to the present disclosure are shown in the accompanying drawings.

[0040] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in the present disclosure can be combined with each other. The technical solution of the present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0041] In related technologies, empirical formulas are relied upon to predict the volume of business opportunities. Such empirical formulas are usually constructed based on limited data accumulated in the past and relatively idealized assumptions. However, in actual business scenarios, the fluctuation of the volume of business opportunities is extremely complex. It is not determined by a simple linear relationship between a single or a few factors, but is affected by the joint action of many nonlinear factors. Since it is difficult for empirical formulas to accurately consider and adapt to these complex nonlinear factors, it is impossible to fully and deeply characterize the true relationship between the volume of business opportunities and many influencing factors. Therefore, the volume of business opportunities predicted by empirical formulas often deviates greatly from the actual situation and has a low accuracy rate. This greatly limits the effectiveness of enterprises in making reasonable decisions and resource planning based on the prediction results. To this end, the present disclosure proposes a method for predicting the volume of business opportunities.

[0042] The business opportunity forecasting method disclosed herein can be used to automatically forecast business opportunity volume using electronic devices that acquire time series data. In the present disclosure, electronic devices include, but are not limited to, mobile phones, tablet computers, laptop computers, personal computers, wearable devices, and ATMs.

[0043] For the convenience of description and to make the technical solutions of the specific embodiments of the present disclosure easier to understand, before describing the business opportunity volume prediction method implemented in the present disclosure, the technical terms involved in the specific embodiments of the present disclosure are explained as follows: Business opportunity volume refers to the number of business opportunities that appear within a certain period.

[0044] A business opportunity refers to an opportunity that has the potential to be converted into an actual transaction. For example, every contact with a real estate agent can be considered a business opportunity.

[0045] A time series refers to a sequence of values ​​of the same statistical indicator arranged in the order of their occurrence.

[0046] Time series features refer to information used to characterize the characteristics of data changing over time.

[0047] Dilated Causal Convolution: A convolution operation that expands the receptive field to capture long-term dependencies without losing temporal order.

[0048] Channel attention weight is used to adjust the importance of features in the channel dimension and highlight the weight value of key channel information.

[0049] Spatial attention weight is used to adjust the importance of features in the feature space and enhance the weight value of local feature perception.

[0050] Figure 1 FIG. 1 shows a schematic diagram of the overall process of a business opportunity prediction method M100 according to an embodiment of the present disclosure. Figure 1 The method shown includes steps S110 to S140, wherein the method can be executed by electronic devices such as a server, a mobile phone, and a computer.

[0051] Specifically, Figure 1 The methods shown include: S110: Acquire time series data, wherein the time series data is used to represent multiple target display data of an application and multiple target user operation data on display content of the application on each of multiple dates; Exemplarily, the target display data may include the exposure of the target display content of the application and / or other display-related data. The exposure corresponding to different target display contents may be regarded as different types of target display data. The target user operation data may include one or more of the total number of clicks, total number of favorites, total number of searches, and total number of filters for the same target display content by all users. The total number of clicks, total number of favorites, total number of searches, or total number of filters corresponding to different target display contents may be regarded as different types of target user operation data. The target display content may be display content related to the amount of business opportunities, such as housing information, etc. Display content unrelated to the amount of business opportunities may include account information pages, APP settings pages, etc.

[0052] Since time series data includes a variety of target display data and a variety of user operation data, it fully covers the information related to the amount of business opportunities during the operation of the application. Therefore, obtaining time series data lays a data foundation for accurately predicting the amount of business opportunities.

[0053] S120, extracting features from the time series data to obtain a first time series feature; For example, data in the form of time series contains a large amount of information reflecting user behavior patterns and market trends. If these time series data cannot be effectively processed and analyzed, it will be difficult to accurately grasp business opportunities, causing companies to miss development opportunities in a highly competitive market. Feature extraction can effectively extract the characteristics or attributes of the time series data that change over time (i.e., the first time series feature), thereby providing data support for predicting the amount of business opportunities.

[0054] S130, adding the first time series feature to the time series data to obtain first data; For example, the dimension of the first time series feature may be the same as the dimension of the time series data, and then the first time series feature and the time series data of the same dimension may be added together, for example, the value at a certain position in the first time series feature may be summed with the value at the same position in the time series data.

[0055] Exemplarily, the result obtained by adding the first time series feature to the time series data can be directly used as the first data, or the result obtained by adding can be processed in a certain way according to actual needs to obtain the first data. For example, feature extraction is performed on the result obtained by adding to obtain a new time series feature, and the result obtained by adding the new time series feature to the time series data can be directly used as the first data. This is not limited here.

[0056] S140: Perform business opportunity quantity prediction based on the first data to obtain a business opportunity quantity prediction result.

[0057] Exemplarily, the first data can be input into a trained deep learning model, such as an Informer model, a Transformer model, etc., and business opportunity quantity prediction can be performed through the deep learning model to obtain a business opportunity quantity prediction result.

[0058] Because opportunity forecasts are based on application data, they can be used to guide application updates or adjustments. For example, if you adjust the placement of targeted content within an application and notice a decrease in lead volume, you can reposition the content elsewhere or restore it to its original position. Furthermore, opportunity forecasts can be used to guide annual business goals.

[0059] The business opportunity volume prediction method of the disclosed embodiment extracts features from time series data to obtain a first time series feature, adds the first time series feature to the time series data to obtain first data, and performs business opportunity volume prediction based on the first data to obtain a business opportunity volume prediction result. Thus, the first time series feature is further added on the basis of the time series data, and while retaining the original information in the time series data, the obtained first data can better characterize the correlation information between the data, thereby enabling the complex relationship between the data to be more accurately captured during the process of business opportunity volume prediction based on the first data, thereby improving the accuracy of the business opportunity volume prediction result obtained based on the first data, and improving the effectiveness of the enterprise in making reasonable decisions and resource planning based on the business opportunity volume prediction result.

[0060] Regarding step S120, in some embodiments of the present disclosure, it may include the following: Figure 2 Steps S121 to S124 are shown.

[0061] S121. Perform a first dilated causal convolution on the time series data to obtain a first convolution result.

[0062] The first dilated causal convolution combines the time order guarantee of causal convolution with the receptive field expansion of dilated convolution, thereby effectively expanding the receptive field of convolution without losing the time series order information, capturing longer-term temporal dependencies, and improving the quality of feature representation.

[0063] Please combine Figure 3 In one example, the dilation factor d in the dilated causal convolutional layer increases exponentially by 2 layer by layer. For example, the second layer extracts features from the first layer with d=1, the third layer extracts features from the second layer with d=2, and the fourth layer extracts features from the third layer with d=4.

[0064] S122. Perform a first weight normalization on the first convolution result to obtain a first normalized result.

[0065] The first weight normalization makes the data features more stable, which is beneficial to improving the training effect and generalization ability of the model. For example, the first weight normalization is to perform normalization on the dimension of weight value.

[0066] S123: Perform a first nonlinear processing on the first normalized result to obtain a first processing result.

[0067] The first nonlinear processing increases the complexity and expressiveness of the model, enabling better fitting of complex patterns in the data. For example, the first nonlinear processing can be performed on the first normalized result using an existing activation function, such as a ReLU (Rectified Linear Unit) function.

[0068] S124: Obtain a first time series feature based on the first processing result.

[0069] As a possible implementation, the first processing result is directly used as the first time series feature.

[0070] As another possible implementation, a first regularization (dropout) process is performed on the first processing result, and the processed first processing result is used as the first time series feature, so as to prevent overfitting.

[0071] As another possible implementation, a first regularization process is performed on the first processing result, and the processed first processing result is again subjected to dilated causal convolution, weight normalization, nonlinear processing and regularization processing, so that the processed result is used as the first time series feature to improve the quality of the first time series feature.

[0072] The business opportunity volume forecasting method of the above embodiment performs a first dilated causal convolution on the time series data to obtain a first convolution result; performs a first weight normalization on the first convolution result to obtain a first normalized result; performs a first nonlinear processing on the first normalized result to obtain a first processed result; and obtains a first time series feature based on the first processed result. The first time series feature obtained through this series of steps can more accurately reflect the inherent patterns in the time series data, provide high-quality feature input for subsequent business opportunity volume forecasting, and help improve the accuracy of business opportunity volume forecasting.

[0073] Regarding step S124, in some embodiments of the present disclosure, it may include the following Figure 4 Steps S1241 to S1244 are shown.

[0074] S1241. Perform a second dilated causal convolution on the first processing result to obtain a second convolution result.

[0075] S1242. Perform a second weight normalization on the second convolution result to obtain a second normalized result.

[0076] S1243: Perform a second nonlinear processing on the second normalized result to obtain a second processing result.

[0077] For the content related to the second dilated causal convolution, the second weight normalization and the second nonlinear processing, reference can be made to the description of the first dilated causal convolution, the first weight normalization and the first nonlinear processing. For the sake of brevity, they are not repeated here.

[0078] S1244. Obtain a first time series feature based on the second processing result.

[0079] As a possible implementation, the second processing result is directly used as the first time series feature.

[0080] As another possible implementation, a second regularization (dropout) process is performed on the second processing result, and the processed second processing result is used as the first time series feature, thereby preventing overfitting.

[0081] The business opportunity volume prediction method of the above-mentioned embodiment adds the first time series feature to the time series data to obtain an addition result; performs a second dilated causal convolution on the addition result to obtain a second convolution result; performs a second weight normalization on the second convolution result to obtain a second normalized result; performs a second nonlinear processing on the second normalized result to obtain a second processing result; and obtains the first data based on the second processing result. Therefore, after adding the first time series feature to the time series data, the second dilated causal convolution, the second weight normalization, and the second nonlinear processing are performed, which further explores the deep features in the data, enhances the hierarchy and richness of the feature representation, helps to more accurately capture the complex patterns in the data, and makes the generated first data contain richer and more valuable information, which can better adapt to complex business opportunity volume prediction tasks, thereby improving the accuracy of business opportunity volume prediction.

[0082] Regarding step S140, in some embodiments of the present disclosure, it may include the following: Figure 5 Steps S141 to S144 are shown.

[0083] S141. Encode the first data based on the self-attention mechanism to obtain encoding features.

[0084] For example, the first data may be encoded based on an encoder of an existing informer model to obtain encoded features. The encoder of the informer model can extract preliminary low-dimensional features and provide efficient feature representation.

[0085] S142. Based on the self-attention mechanism, the encoded features are decoded to obtain decoded features.

[0086] For example, the encoded features may be encoded based on a decoder of an existing Informer model to obtain the decoded features.

[0087] S143. Determine target features based on the decoded features.

[0088] As a possible implementation method, considering that the decoded features can already represent key information related to the business opportunity volume in the time series data, the decoded features can be directly used as target features to provide data support for business opportunity volume prediction.

[0089] As another possible implementation, the decoded features may be further processed to further enrich key information related to the business opportunity volume, and the processed decoded features may be used as target features to further improve the accuracy of business opportunity volume prediction.

[0090] S144. Predict the business opportunity quantity based on the target features to obtain a business opportunity quantity prediction result.

[0091] Exemplarily, the target feature may be fully connected to obtain a fully connected feature, and the fully connected feature may be mapped using a preset activation function, and the mapping result may be used as a business opportunity prediction result.

[0092] The business opportunity volume prediction method of the above-described embodiment encodes the first data based on the self-attention mechanism to obtain encoded features; decodes the encoded features based on the self-attention mechanism to obtain decoded features; determines target features based on the decoded features; and predicts business opportunity volume based on the target features to obtain a business opportunity volume prediction result. Thus, encoding and decoding the first data based on the self-attention mechanism can automatically learn the dependencies between elements at different positions in the first data, effectively capturing the global features of the first data. By determining the target features and predicting business opportunity volume based on them, key information related to business opportunity volume can be highlighted, interference factors can be eliminated, and the business opportunity volume prediction result can be made more accurate and reliable.

[0093] Regarding step S143, as a possible implementation, it can be specifically: capturing the temporal dependency in the decoding features through a long short-term memory network (LSTM), and using the captured second temporal feature as the target feature.

[0094] The structure and parameters of the long short-term memory network are not limited in this disclosure.

[0095] The business opportunity forecasting method described in the above embodiment uses a long short-term memory network to capture temporal dependencies in decoded features. This effectively addresses long-term dependencies in time series data and avoids information loss. Using the captured second temporal feature as the target feature allows the target feature to fully reflect the temporal changes in the data and establish a closer connection with the business opportunity volume, thereby improving the accuracy and stability of the business opportunity forecast.

[0096] Regarding step S143, as another possible implementation, it may include the following: Figure 6 Steps S1431 to S1434 are shown.

[0097] S1431. Extract the first channel attention weight from the decoded features.

[0098] The first channel attention weight can be extracted from the decoding features based on an existing channel attention weight extraction module or a channel attention weight extraction algorithm.

[0099] S1432. Perform weighted calculation on the decoding features based on the first channel attention weight to obtain the first channel attention features.

[0100] Exemplarily, the first channel attention weight can be weighted to the corresponding channel of the decoded feature by multiplication to obtain the first channel attention feature.

[0101] S1433. Extract the first spatial attention weight from the first channel attention feature.

[0102] The first spatial attention weight can be extracted from the first channel attention feature based on an existing spatial attention weight extraction module or a spatial attention weight extraction algorithm.

[0103] S1434. Perform weighted calculation on the first channel attention feature based on the first spatial attention weight, and use the calculation result as the target feature.

[0104] Exemplarily, the first spatial attention weight can be weighted to the corresponding space of the first channel attention feature by multiplication to obtain the first channel attention feature.

[0105] The business opportunity prediction method of the above-mentioned embodiment extracts the first channel attention weight from the decoded features; performs weighted calculation on the decoded features based on the first channel attention weight to obtain the first channel attention feature; extracts the first spatial attention weight from the first channel attention feature; and performs weighted calculation on the first channel attention feature based on the first spatial attention weight, and uses the calculation result as the target feature. Therefore, by extracting the first channel attention weight and the first spatial attention weight, and performing weighted calculation on the decoded features based on them, the model can pay more attention to important channel and spatial location information related to the business opportunity volume, reduce the interference of irrelevant information, make the generated target feature more targeted and effective, and thus improve the accuracy and reliability of the business opportunity volume prediction.

[0106] Regarding step S1431, in some embodiments of the present disclosure, it may include the following: Figure 7 Steps S311 to S314 are shown.

[0107] S311. Perform global average pooling and global maximum pooling on the decoded features respectively.

[0108] S312. Process the global average pooling result and the global maximum pooling result respectively through a shared multilayer perceptron (MLP).

[0109] S313: The global average pooling result after the processing and the global maximum pooling result after the processing are fused to obtain a fusion feature.

[0110] For example, the processed global average pooling result and the processed global maximum pooling result may be added, and the added result may be used as a fusion feature.

[0111] S314. Map the fused features through a first activation function to obtain a first channel attention weight.

[0112] The first activation function can be set according to actual needs. The first channel attention weight refers to the channel attention weight of all channels in the decoded features, rather than the channel attention weight of a specific channel. The same applies to the second channel attention weight, the first spatial attention weight, and the second spatial attention weight. In an example, the visualization diagram of steps S311 to S314 is as follows: Figure 8 shown.

[0113] The business opportunity prediction method described in the above embodiment performs global average pooling and global maximum pooling on the decoded features, respectively, and fuses them through a shared multi-layer perceptron. This fully extracts global information from the decoded features, combining the advantages of different statistical methods to produce richer fused features. Mapping the fused features using a first activation function to obtain first-channel attention weights can adaptively adjust feature weights along the channel dimension, highlighting channel information that has a significant impact on business opportunity prediction, improving the model's ability to mine features, and providing more effective first-channel attention weight allocation for accurate business opportunity prediction.

[0114] Regarding step S1433, in some embodiments of the present disclosure, it may include the following: Figure 9 Steps S331 to S334 are shown.

[0115] S331. Perform global average pooling and global maximum pooling on the attention features of the first channel along the feature point channel direction.

[0116] S332. Concatenate the global average pooling result and the global maximum pooling result of the same channel to obtain a concatenated feature.

[0117] S333. Perform convolution on the splicing features.

[0118] S334. Map the convolutional concatenation features through a second activation function to obtain a first spatial attention weight.

[0119] The second activation function can be set according to actual needs. In one example, the visualization example diagram of steps S331 to S334 is as follows: Figure 10 shown.

[0120] The business opportunity prediction method described above performs global average pooling and global maximum pooling on the first channel attention features along the feature point channel direction, then concatenates, convolves, and maps them through a second activation function to obtain the first spatial attention weight. This further filters and strengthens the features from a spatial dimension. This approach allows the model to focus on key spatial regions related to business opportunity volume, enhances the model's perception of spatial features, and improves the quality of target features, thereby improving the accuracy of business opportunity prediction and the efficiency of utilizing spatial information.

[0121] Regarding step S143, as another possible implementation, it may include the following: Figure 11 Steps S1431' to S1435' are shown.

[0122] S1431', capturing the temporal dependency in the decoding features through a long short-term memory network to obtain a second temporal feature.

[0123] The structure and parameters of the long short-term memory network are not limited in this disclosure.

[0124] S1432', extract the second channel attention weight from the second temporal feature.

[0125] S1433′: perform weighted calculation on the second temporal feature based on the second channel attention weight to obtain the second channel attention feature.

[0126] S1434', extract the second spatial attention weight from the second channel attention feature.

[0127] S1435', perform weighted calculation on the second channel attention feature based on the second spatial attention weight, and use the calculation result as the target feature.

[0128] For details related to steps S1432' to S1435', refer to the description of steps S1431 to S1434, and for the sake of brevity, they will not be repeated here.

[0129] The business opportunity forecasting method described in the above embodiment combines the long short-term memory network to capture temporal dependencies with the channel attention mechanism and the spatial attention mechanism to comprehensively mine the temporal and spatial information in the decoded features. This method not only considers the temporal variation of data but also highlights key channel and spatial location information. This allows the generated target features to more comprehensively and accurately reflect information related to business opportunity volume, effectively improving the accuracy of business opportunity forecasts.

[0130] In some embodiments of the present disclosure, before step S110, the following steps may also be included: Figure 12 Steps S150 to S170 are shown.

[0131] S150. Acquire different types of target data stored on each of a plurality of dates, where the different types of target data include different types of target display data of an application and different types of target user operation data for display content of the application. The storage period of the target data is less than or equal to one day.

[0132] S160 . For each date among the multiple dates, for each type of target data, summing up the target data of the same type stored multiple times on the date.

[0133] Because different types of target data may have different storage periods, the data volumes of different types of target data within a statistical period may differ. For example, if type A target data is stored three times, it will have three data values, while type B target data is stored eight times, it will have eight data values. This results in the inability to generate valid time series data, and thus, the inability to accurately predict the volume of business opportunities. In the disclosed embodiment, however, the same type of target data stored multiple times on each date is summed up on a daily basis, ensuring that the data volumes of different types of target data within each statistical period (one day is one statistical period) are the same, each with one data value. This allows the generation of time series data with aligned data volumes, facilitating accurate business opportunity prediction.

[0134] In an example, type A target data stores 3 data values ​​a day, and type B target data stores 8 data values ​​a day. The sum of the 3 data values ​​of type A target data can be used as the type A target data for that day, and the sum of the 8 data values ​​of type B target data can be used as the type B target data for that day.

[0135] It should be noted that the specific numerical values ​​mentioned above are only used as examples to illustrate the implementation of the present disclosure in detail and should not be understood as limiting the present disclosure. In other examples, implementation methods, or embodiments, other numerical values ​​can be selected according to the present disclosure and are not specifically limited here.

[0136] S170 , based on the order of dates, generate time series data according to the sum of each type of target data on each date among multiple dates.

[0137] Exemplarily, the time series data may be in the form of a matrix, where each row of the matrix represents a type of target data, each column of the matrix represents a different type of target data on the same date, and the date gradually moves backward as the number of columns increases.

[0138] The business opportunity volume prediction method of the above-mentioned embodiment, before obtaining the time series data, sums the target data of the same type stored multiple times on each date, and generates time series data in chronological order of dates, thereby ensuring the consistency and integrity of the time series data, reducing data redundancy, highlighting the time trend of the data, and providing high-quality basic data for subsequent feature extraction and prediction, which is conducive to improving the subsequent data processing efficiency and the accuracy of the business opportunity volume prediction results.

[0139] In some embodiments of the present disclosure, before step S150, the following steps may also be included: Figure 13 Steps S180 to S200 are shown.

[0140] S180: Acquire different types of data and business opportunity quantities stored on each of a plurality of dates, where the different types of data include different types of display data of the application and different types of user operation data on display content of the application.

[0141] S190 . Calculate the correlation coefficient between each type of data and the amount of business opportunities stored on each of the multiple dates.

[0142] For example, for each date among multiple dates, for each type of data, the same type of data stored multiple times on the date can be summed up to obtain the data of that type on the date, and the amount of business opportunities stored multiple times on the date can be summed up to obtain the amount of business opportunities on the date. Then, the Spearman correlation coefficient, Pearson correlation coefficient or other correlation coefficient calculation methods can be used to calculate the correlation coefficient between each type of data and the amount of business opportunities in multiple dates.

[0143] S200 , taking data with a correlation coefficient greater than or equal to a correlation coefficient threshold as target data.

[0144] The correlation coefficient threshold can be set as needed. Data with a correlation coefficient greater than or equal to the threshold has a high correlation with the number of business opportunities and can be used as target data for business opportunity forecasting. Data with a correlation coefficient less than the threshold has a low correlation with the number of business opportunities, which not only prolongs the forecasting time but also affects the accuracy of the forecast results. Therefore, data with a correlation coefficient less than the threshold can be excluded, meaning that this data is not used for business opportunity forecasting.

[0145] The business opportunity volume forecasting method of the above-described embodiment, before acquiring target data, calculates the correlation coefficient between different types of data and business opportunity volume, and selects data with a correlation coefficient greater than or equal to a correlation coefficient threshold as target data. This method can specifically screen out data closely related to business opportunity volume. This data screening method can reduce the interference of irrelevant data on business opportunity volume forecasting, reduce the complexity of data processing during the business opportunity volume forecasting process, improve data quality, and enable subsequent business opportunity volume forecasting to focus more on the analysis and learning of data closely related to business opportunity volume, thereby improving the accuracy and efficiency of business opportunity volume forecasting and saving computing resources and time costs.

[0146] Please combine Figure 14In one example, the business opportunity prediction model is improved based on the Informer model and can be used to implement steps S120 to S140 above. Specifically, a Temporal Convolutional Network (TCN) module is introduced at the top layer of the Informer model, and an LSTM module and a Convolutional Block Attention Module (CBAM) are introduced at the bottom layer of the Informer model. The TCN module, Informer encoder, Informer decoder, LSTM module, CBAM module, and output module are connected in series.

[0147] Furthermore, please combine Figure 15 The TCN module includes a first residual unit and a second residual unit connected in series. The first residual unit and the second residual unit, respectively, include a first dilated causal convolutional layer, a first weight normalization layer, a first activation function layer, a first random dropout layer, a second dilated causal convolutional layer, a second weight normalization layer, a second activation function layer, and a second random dropout layer. The first dilated causal convolutional layer, the first weight normalization layer, the first activation function layer, the first random dropout layer, the second dilated causal convolutional layer, the second weight normalization layer, the second activation function layer, and the second random dropout layer are connected in series. Furthermore, a residual connection is established between the input of the first dilated causal convolutional layer and the output of the second random dropout layer via a 1×1 convolution. This not only helps alleviate the vanishing gradient problem but also promotes effective feature transfer. The dilation factor d in the first and second dilated causal convolutional layers can be increased layer by layer in an exponential manner of 2 (e.g., 1, 2, 4, 8) to expand the receptive field. The activation functions in the first and second activation function layers can be pre-set based on actual needs, for example, as ReLU functions. The first weight normalization layer, the second weight normalization layer, the first random dropout layer and the second random dropout layer can be set according to specific task requirements.

[0148] The TCN module can be used to implement steps S120 and S130 above. In the first residual unit, a first dilated causal convolution layer is used to implement step S121 above, a first weight normalization layer is used to implement step S122 above, a first activation function layer is used to implement step S123 above, a first random dropout layer is used to perform a first regularization process on the first processing result, a second dilated causal convolution layer is used to implement step S1241 above, a second weight normalization layer is used to implement step S1242 above, a second activation function layer is used to implement step S1243 above, and a second random dropout layer is used to implement step S1244 above. A residual connection is used to add the first time series feature to the time series data to obtain a first addition result. The second residual unit is used to further process the first addition result output by the first residual unit to obtain a second addition result (i.e., the first data). For details related to the second residual unit, please refer to the implementation of the first residual unit above and will not be repeated here for the sake of brevity.

[0149] As a temporal attention module, the TCN module can obtain the temporal correlation of various types of data in time series data, achieve cross-domain feature alignment, reduce domain distribution differences, enhance model migration capabilities, and make it easier for the Informer model to learn the correlation information between various types of data in time series data.

[0150] The Informer model can capture information over long timescales. The Informer model may include a serially connected encoder and decoder, where the encoder may include a multi-head probabilistic sparse self-attention layer, and the decoder may include a masked multi-head probabilistic sparse self-attention layer. The encoder is used to implement step S141 above, and the decoder is used to implement step S142 above.

[0151] The LSTM module can deeply model temporal features, capture long-term dependencies, and predict dynamic changes. The LSTM module can be used to implement the above step S1431'.

[0152] Please combine Figure 16The CBAM module can include a channel attention module (CAM), a first operation unit, a spatial attention module (SAM), and a second operation unit. The channel attention module can be used to implement the above step S1432', the first operation unit can be used to implement the above step S1433', the spatial attention module can be used to implement the above step S1434', and the second operation unit can be used to implement the above step S1435'. The CBAM module combines channel attention and spatial attention. The two attention mechanisms have different focuses and complement each other. Channel attention enables the model to re-evaluate the contribution of each channel as a whole, while spatial attention redistributes importance locally. In this way, the output of the LSTM module is finely optimized, and the improved model structure can achieve better prediction results.

[0153] The output module may include a fully connected layer and a mapping layer. The fully connected layer is used to perform fully connected processing on the target features to obtain fully connected features. The mapping layer is used to map the fully connected features through a preset activation function. The mapping results can be used as business opportunity prediction results and output.

[0154] The training set of the above-mentioned business opportunity volume prediction model may include time series data with labeled business opportunity volume. During the training process of the business opportunity volume prediction model, the time series data in the training set can be input into the business opportunity volume prediction model to predict the business opportunity volume and obtain the predicted business opportunity volume. The loss value can be calculated based on the predicted business opportunity volume and the labeled business opportunity volume, and the business opportunity volume prediction model can be adjusted based on the loss value to obtain the trained business opportunity volume prediction model.

[0155] Please combine Figure 17 In an example, the business opportunity volume prediction method implemented based on the above-mentioned business opportunity volume prediction model may include the following steps S401 to S404. The contents related to steps S401 to S404 can refer to the description of the above-mentioned implementation method. For the sake of brevity, they will not be repeated here.

[0156] In step S401, different types of target data stored on each of multiple dates are obtained, where the different types of target data include different types of target display data of the application and different types of target user operation data for the display content of the application. The storage period of the target data is less than or equal to one day.

[0157] In step S402 , for each date among the multiple dates and for each type of target data, the target data of the same type stored multiple times on the date are summed up.

[0158] In step S403 , based on the order of dates, time series data is generated according to the sum of each type of target data on each date in the multiple dates.

[0159] In step S404, the time series data is input into the business opportunity quantity prediction model to perform business opportunity quantity prediction and obtain a business opportunity quantity prediction result.

[0160] Based on any of the above embodiments, the present disclosure also provides a business opportunity quantity prediction device.

[0161] Figure 18 It is a schematic block diagram of the structure of a business opportunity volume prediction device according to an embodiment of the present disclosure.

[0162] like Figure 18 As shown, the business opportunity prediction device includes: An acquisition module 110 is configured to acquire time series data, wherein the time series data is used to represent various target display data of an application and various target user operation data on display content of the application on each of a plurality of dates; A feature extraction module 120 is used to extract features from the time series data to obtain a first time series feature; A processing module 130 is configured to add the first time series feature to the time series data to obtain first data; The prediction module 140 is configured to perform business opportunity quantity prediction based on the first data to obtain a business opportunity quantity prediction result.

[0163] The business opportunity volume prediction device may be in the form of computer software, and each module of the business opportunity volume prediction device may be implemented by a computer software module.

[0164] In some embodiments of the present disclosure, the feature extraction module 120 is used to: perform a first dilated causal convolution on the time series data to obtain a first convolution result; perform a first weight normalization on the first convolution result to obtain a first normalized result; perform a first nonlinear processing on the first normalized result to obtain a first processing result; and obtain a first time series feature based on the first processing result.

[0165] In some embodiments of the present disclosure, the feature extraction module 120 is used to: perform a second dilated causal convolution on the first processing result to obtain a second convolution result; perform a second weight normalization on the second convolution result to obtain a second normalized result; perform a second nonlinear processing on the second normalized result to obtain a second processing result; and obtain a first time series feature based on the second processing result.

[0166] In some embodiments of the present disclosure, the prediction module 140 is used to: encode the first data based on the self-attention mechanism to obtain encoding features; decode the encoding features based on the self-attention mechanism to obtain decoding features; determine the target features based on the decoding features; and predict the business opportunity volume based on the target features to obtain a business opportunity volume prediction result.

[0167] In some embodiments of the present disclosure, the prediction module 140 is configured to capture temporal dependencies in decoding features through a long short-term memory network, and use the captured second temporal features as target features.

[0168] In some embodiments of the present disclosure, the prediction module 140 is used to: extract a first channel attention weight from the decoded feature; perform weighted calculation on the decoded feature based on the first channel attention weight to obtain a first channel attention feature; extract a first spatial attention weight from the first channel attention feature; and perform weighted calculation on the first channel attention feature based on the first spatial attention weight, and use the calculation result as the target feature.

[0169] In some embodiments of the present disclosure, the prediction module 140 is used to: perform global average pooling and global maximum pooling on the decoded features respectively; process the global average pooling results and the global maximum pooling results respectively through a shared multi-layer perceptron; fuse the processed global average pooling results and the processed global maximum pooling results to obtain fused features; and map the fused features through a first activation function to obtain a first channel attention weight.

[0170] In some embodiments of the present disclosure, the prediction module 140 is used to: perform global average pooling and global maximum pooling on the first channel attention features along the feature point channel direction; splice the global average pooling results and the global maximum pooling results of the same channel to obtain spliced ​​features; convolve the spliced ​​features; and map the convolved spliced ​​features through a second activation function to obtain a first spatial attention weight.

[0171] In some embodiments of the present disclosure, the prediction module 140 is used to: capture the temporal dependency in the decoding features through a long short-term memory network to obtain a second temporal feature; extract a second channel attention weight from the second temporal feature; perform weighted calculation on the second temporal feature based on the second channel attention weight to obtain a second channel attention feature; extract a second spatial attention weight from the second channel attention feature; and perform weighted calculation on the second channel attention feature based on the second spatial attention weight, and use the calculation result as the target feature.

[0172] In some embodiments of the present disclosure, the business opportunity volume prediction device also includes: a second acquisition module, which is used to acquire different types of target data stored on each of multiple dates, and the different types of target data include different types of target display data of the application and different types of target user operation data for the display content of the application, and the storage period of the target data is less than or equal to one day; a second processing module, which is used to sum up the same type of target data stored multiple times on each date in the multiple dates for each type of target data; and a generation module, which is used to generate time series data based on the order of the dates and according to the results of summing up each type of target data on each date in the multiple dates.

[0173] In some embodiments of the present disclosure, the business opportunity volume prediction device also includes: a third acquisition module, used to obtain different types of data and business opportunity volumes stored on each of multiple dates, the different types of data including different types of display data of the application and different types of user operation data for the display content of the application; a calculation module, used to calculate the correlation coefficient between each type of data and the business opportunity volume based on the different types of data and business opportunity volumes stored on each of the multiple dates; and a determination module, used to take data with a correlation coefficient greater than or equal to a correlation coefficient threshold as target data.

[0174] The implementation process of the functions and effects of each module in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0175] The execution subject of the business opportunity prediction method in the specific embodiment of the present disclosure can be a server, a mobile phone, a computer or other electronic device.

[0176] Therefore, based on any of the above embodiments, the present disclosure further provides an electronic device, which can execute the business opportunity quantity prediction method of any of the above embodiments described in the present disclosure.

[0177] Figure 19 1 is a schematic block diagram of the structure of an electronic device 1000 according to an embodiment of the present disclosure.

[0178] The hardware structure of the electronic device 1000 can be implemented using a bus architecture. The bus architecture can include any number of interconnecting buses and bridges, depending on the specific application and overall design constraints of the hardware. The bus 1100 connects various circuits including one or more processors 1200, memory 1300, and / or hardware modules. The bus 1100 can also connect various other circuits 1400 such as peripheral devices, voltage regulators, power management circuits, external antennas, etc.

[0179] Bus 1100 may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Component Architecture (EISA) bus. Buses can be classified as address buses, data buses, control buses, and the like. For ease of illustration, this figure shows only one connecting line, but this does not imply that there is only one bus or only one type of bus.

[0180] The present disclosure also provides a readable storage medium having a computer program stored therein, which is used to implement the above-mentioned method when the computer program is executed by a processor. "Readable storage medium" can be any device that can contain, store, communicate, propagate or transmit a program for use in an instruction execution system, device or equipment or in combination with these instruction execution systems, devices or equipment. More specific examples of readable storage media include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and editable read-only memory (EPROM or flash memory), an optical fiber device, and a portable read-only memory (CDROM), etc.

[0181] The present disclosure also provides a computer program product. The method of the present disclosure can be implemented in whole or in part using software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed, the process or function of the present disclosure is performed in whole or in part.

[0182] A computer program or instruction can be stored in a readable storage medium or transferred from one readable storage medium to another. For example, the computer program or instruction can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The readable storage medium can be any accessible medium or a data storage device such as a server or data center that integrates one or more accessible media. The accessible medium can be a magnetic medium such as a floppy disk, hard disk, or magnetic tape; an optical medium such as a digital video disk; or a semiconductor medium such as a solid-state drive. The computer-readable storage medium can be a volatile or non-volatile storage medium, or can include both volatile and non-volatile types of storage media.

[0183] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0184] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0185] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0186] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0187] In the description of this specification, the description with reference to the terms "one embodiment / method", "some embodiments / methods", "example", "specific example", or "some examples" means that the specific features, structures, or characteristics described in conjunction with the embodiment / method or example are included in at least one embodiment / method or example of the present disclosure. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment / method or example. Moreover, the specific features, structures, or characteristics described may be combined in a suitable manner in any one or more embodiments / methods or examples. In addition, those skilled in the art may combine and combine different embodiments / methods or examples described in this specification and the features of different embodiments / methods or examples, unless they are contradictory.

[0188] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. Throughout the present disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0189] Those skilled in the art will appreciate that the above embodiments are merely intended to clearly illustrate the present disclosure and are not intended to limit the scope of the present disclosure. Other changes or modifications may be made based on the above disclosure, and such changes or modifications are still within the scope of the present disclosure.

Claims

1. A method for predicting business opportunity volume, characterized in that: include: Acquire time series data, wherein the time series data is used to represent multiple target display data of an application and multiple target user operation data on display content of the application on each of multiple dates; Performing feature extraction on the time series data to obtain a first time series feature; Adding the first time series feature to the time series data to obtain first data; and Business opportunity quantity prediction is performed based on the first data to obtain a business opportunity quantity prediction result.

2. The business opportunity forecasting method according to claim 1, characterized in that: Performing feature extraction on the time series data to obtain a first time series feature includes: Performing a first dilated causal convolution on the time series data to obtain a first convolution result; Performing a first weight normalization on the first convolution result to obtain a first normalized result; performing a first nonlinear processing on the first normalized result to obtain a first processing result; and The first time series feature is obtained based on the first processing result.

3. The business opportunity forecasting method according to claim 1, characterized in that: Performing a business opportunity quantity forecast based on the first data to obtain a business opportunity quantity forecast result includes: Encoding the first data based on a self-attention mechanism to obtain encoding features; Based on the self-attention mechanism, the encoded features are decoded to obtain decoded features; Determining a target feature based on the decoded feature; and Business opportunity quantity prediction is performed based on the target feature to obtain the business opportunity quantity prediction result.

4. The business opportunity forecasting method according to claim 3, characterized in that: Determining a target feature according to the decoded feature includes: The temporal dependency in the decoding feature is captured by a long short-term memory network, and the captured second temporal feature is used as the target feature.

5. The business opportunity forecasting method according to claim 3, characterized in that: Determining a target feature according to the decoded feature includes: Extracting a first channel attention weight from the decoded features; Performing weighted calculation on the decoding feature based on the first channel attention weight to obtain a first channel attention feature; Extracting a first spatial attention weight from the first channel attention feature; and The first channel attention feature is weightedly calculated based on the first spatial attention weight, and the calculation result is used as the target feature.

6. The business opportunity forecasting method according to claim 3, characterized in that: Determining a target feature according to the decoded feature includes: Capturing the temporal dependency in the decoding features through a long short-term memory network to obtain a second temporal feature; Extracting a second channel attention weight from the second temporal feature; Performing weighted calculation on the second temporal feature based on the second channel attention weight to obtain a second channel attention feature; Extracting a second spatial attention weight from the second channel attention feature; and The second channel attention feature is weightedly calculated based on the second spatial attention weight, and the calculation result is used as the target feature.

7. The business opportunity forecasting method according to any one of claims 1 to 6, characterized in that: Before acquiring time series data, also include: Acquire different types of target data stored on each of a plurality of dates, the different types of target data including different types of target display data of the application and different types of target user operation data on display content of the application, wherein the storage period of the target data is less than or equal to one day; For each of the multiple dates, for each type of target data, summing the target data of the same type stored multiple times on the date; and Based on the order of dates, the time series data is generated according to the result of summing up each type of target data on each date among the multiple dates.

8. The business opportunity forecasting method according to claim 7, characterized in that: Before getting the different types of target data stored on each of the multiple dates, it also includes: Acquire different types of data and business opportunity quantities stored on each of a plurality of dates, wherein the different types of data include different types of display data of the application and different types of user operation data on display content of the application; Calculating, based on the different types of data and the amount of business opportunities stored on each of the multiple dates, a correlation coefficient between each type of data and the amount of business opportunities; and The data whose correlation coefficient is greater than or equal to the correlation coefficient threshold is used as the target data.

9. An electronic device, characterized in that: include: a memory storing execution instructions; as well as A processor, wherein the processor executes the execution instructions stored in the memory, so that the processor executes the business opportunity volume prediction method according to any one of claims 1 to 8.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the business opportunity volume prediction method according to any one of claims 1 to 8 is implemented.

Images