User demand analysis method and device based on browsing consumption time series data

Abstract

The present invention provides a method and device for analyzing user demand based on browsing consumption time series data, the method comprising: obtaining historical behavior data of users during a monitoring period, intercepting the behavior data of users within a predetermined period respectively according to the occurrence of transaction behaviors of users in the historical behavior data to construct positive sample data and negative sample data; performing numerical processing on the positive sample data and the negative sample data to generate a numerical sequence, and performing vector processing on the numerical sequence to generate a positive and negative sample set matrix; training and verifying a learning model constructed based on a deep learning long short-term memory neural network through the positive and negative sample set matrix to obtain a demand analysis model; collecting the user's behavior data to be tested, and obtaining user demand according to the behavior data to be tested and the demand analysis model.

Description

Technical Field

[0001] The present invention relates to the field of big data, and in particular to a method and device for analyzing user needs based on browsing consumption time series data. Background Art

[0002] With the rapid development of emerging science and technology represented by big data, cloud computing, artificial intelligence, and their wide application in the financial field, banks are paying more and more attention to the refined development of product marketing. How to further strengthen the effective identification of user needs and user characteristics by using accurate account history data, user browsing records and other information about consumer behavior has become an important issue in the banking industry.

[0003] At present, the Internet has become the most convenient channel for obtaining information. Financial institutions can effectively identify user needs by analyzing user transaction behavior, browsing behavior, click behavior, search behavior and other information, and then intelligently judge the degree of user demand for products based on user needs; by deeply mining user historical search records and click behavior data, finding the correlation pattern of user needs can effectively improve the marketing success rate of product push. For this reason, with the rapid development of informatization, mining and analyzing massive user browsing consumption record data has become a research hotspot, and more and more methods have been proposed to be applied to the analysis and prediction of user demand for products, such as Logistic regression method, decision tree and random forest method; but due to technical difficulties such as large data volume and complex processing, most of the research is still qualitative analysis of influencing factors to construct theoretical models, and the performance is more or less limited. In summary, the industry is in urgent need of a user demand analysis method with high accuracy. Summary of the invention

[0004] The purpose of the present invention is to provide a user demand analysis method and device based on browsing consumption time series data, to provide user demand analysis with high accuracy and significant effect, and to provide effective data support for subsequent research.

[0005] To achieve the above-mentioned purpose, the user demand analysis method based on browsing consumption time series data provided by the present invention specifically includes: obtaining the historical behavior data of users during the monitoring period, and constructing positive sample data and negative sample data by intercepting the behavior data of users within a predetermined period according to the transaction behavior of users in the historical behavior data; digitizing the positive sample data and the negative sample data to generate a numerical sequence, and vectorizing the numerical sequence to generate a positive and negative sample set matrix; training and verifying the learning model constructed based on a deep learning long short-term memory neural network through the positive and negative sample set matrix to obtain a demand analysis model; collecting the user's behavior data to be tested, and obtaining user needs based on the behavior data to be tested and the demand analysis model.

[0006] In the above-mentioned user demand analysis method based on browsing consumption time series data, preferably, according to the transaction behavior of the user in the historical behavior data, the behavior data of the user within the predetermined period is respectively intercepted to construct positive sample data and negative sample data, including: when the historical behavior data contains transaction behavior, according to the transaction occurrence time of the transaction behavior, the first behavior data within the predetermined first period before the transaction occurrence time is intercepted, and the positive sample data is generated according to the first behavior data; when the user has no transaction behavior within the predetermined second period in the historical behavior data, the second behavior data of the user within the predetermined third period is intercepted, and the negative sample data is generated according to the second behavior data.

[0007] In the above-mentioned user demand analysis method based on browsing consumption time series data, preferably, when the historical behavior data includes transaction behavior, the first behavior data within a predetermined first period before the transaction occurrence time is intercepted according to the transaction occurrence time of the transaction behavior, including: when the historical behavior data includes multiple transaction behaviors, the transaction occurrence time of the Nth transaction behavior is used as the starting time; after pushing forward the predetermined fourth period with the starting time as the starting point, the first behavior data within the predetermined first period is intercepted.

[0008] In the above-mentioned user demand analysis method based on browsing consumption time series data, preferably, when the user has not performed any transaction behavior within a predetermined second period in the historical behavior data, the second behavior data of the user within a predetermined third period is intercepted, including: the third period is smaller than the second period, and the start time and the end time are different.

[0009] In the above-mentioned user demand analysis method based on browsing consumption time series data, preferably, before the positive sample data and the negative sample data are numerically processed to generate a numerical sequence, it also includes: obtaining the URL browsing records in the positive sample data and the negative sample data, eliminating irrelevant URLs according to the correlation between the URL browsing records and the transaction behavior, and obtaining associated URL strings.

[0010] In the above-mentioned user demand analysis method based on browsing consumption time series data, preferably, the numerical processing of the positive sample data and the negative sample data to generate a numerical sequence includes: numbering the associated URL strings, constructing an association dictionary of URL strings and URL numbers; and generating a numerical sequence based on the association dictionary and the associated URL strings.

[0011] In the above-mentioned user demand analysis method based on browsing consumption time series data, preferably, vectorizing the numerical sequence to generate a positive and negative sample set matrix includes: mapping the numerical sequence into a multidimensional vector to obtain a positive and negative sample set matrix according to a preset unit browsing time.

[0012] In the above-mentioned user demand analysis method based on browsing consumption time series data, preferably, vectorizing the numerical sequence to generate a positive and negative sample set matrix includes: obtaining the user account balance change data in the positive sample data and the negative sample data, and constructing a unit time amount change vector based on the user account balance change data with a preset time length as a unit; and aggregating the unit time amount change vector to generate a positive and negative sample set matrix.

[0013] In the above-mentioned user demand analysis method based on browsing consumption time series data, preferably, obtaining the demand analysis model by training and verifying the learning model constructed based on the deep learning long short-term memory neural network through the positive and negative sample set matrix includes: dividing the positive and negative sample set matrix into a training set and a test set according to a preset ratio; training the learning model through the training set, and verifying the trained learning model through the test set to obtain the demand analysis model.

[0014] The present invention also provides a user demand analysis device based on browsing consumption time series data, the device comprising a sample collection module, a preprocessing module, a model building module and an analysis module; the sample collection module is used to obtain the historical behavior data of users during the monitoring period, and according to the transaction behavior of users in the historical behavior data, the behavior data of users within a predetermined period are respectively intercepted to construct positive sample data and negative sample data; the preprocessing module is used to perform numerical processing on the positive sample data and the negative sample data to generate a numerical sequence, and perform vector processing on the numerical sequence to generate a positive and negative sample set matrix; the model building module is used to train and verify the learning model constructed based on the deep learning long short-term memory neural network through the positive and negative sample set matrix to obtain a demand analysis model; the analysis module is used to collect the user's behavior data to be tested, and obtain user demand according to the behavior data to be tested and the demand analysis model.

[0015] In the above-mentioned user demand analysis device based on browsing consumption time series data, preferably, the sample collection module includes an extraction unit, and the extraction unit is used for, when the historical behavior data includes transaction behavior, intercepting the first behavior data within a predetermined first period before the transaction occurrence time according to the transaction occurrence time of the transaction behavior, and generating positive sample data according to the first behavior data; and, when the user has not performed transaction behavior within a predetermined second period in the historical behavior data, intercepting the second behavior data of the user within a predetermined third period, and generating negative sample data according to the second behavior data.

[0016] In the above-mentioned user demand analysis device based on browsing consumption time series data, preferably, the preprocessing module also includes a screening unit, and the screening unit is used to obtain the URL browsing records in the positive sample data and the negative sample data, eliminate irrelevant URLs according to the correlation between the URL browsing records and the transaction behavior, and obtain associated URL strings.

[0017] In the above-mentioned user demand analysis device based on browsing consumption time series data, preferably, the preprocessing module includes a sequence unit and a vector unit; the sequence unit is used to number the associated URL strings, construct an association dictionary of URL strings and URL numbers, and generate a numerical sequence according to the association dictionary and the associated URL strings; the vector unit is used to map the numerical sequence into a multidimensional vector to obtain a positive and negative sample set matrix according to a preset unit browsing time.

[0018] In the above-mentioned user demand analysis device based on browsing consumption time series data, preferably, the preprocessing module includes: obtaining the user account balance change data in the positive sample data and the negative sample data, and constructing a unit time amount change vector based on the preset time length according to the user account balance change data; aggregating the unit time amount change vector to generate a positive and negative sample set matrix.

[0019] In the above-mentioned user demand analysis device based on browsing consumption time series data, preferably, the model construction module includes: dividing the positive and negative sample set matrix into a training set and a test set according to a preset ratio; training the learning model through the training set, and verifying the trained learning model through the test set to obtain a demand analysis model.

[0020] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the computer program.

[0021] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for executing the above method.

[0022] The beneficial technical effect of the present invention is that it accurately analyzes user needs through the behavioral patterns contained in time series data such as massive user browsing records and historical consumption information, combined with a long short-term memory neural network deep model; it can not only process browsing consumption time series data, but also effectively prevent convergence problems caused by gradient disappearance or explosion, and the analysis effect is also very significant. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of the present application, and do not constitute a limitation of the present invention. In the drawings:

[0024] Figure 1 A flowchart of a method for analyzing user needs based on browsing consumption time series data provided by an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of a process for obtaining positive and negative sample data provided by an embodiment of the present invention;

[0026] Figure 3 A schematic diagram of a positive sample selection rule provided by an embodiment of the present invention;

[0027] Figure 4 A schematic diagram of a negative sample selection rule provided by an embodiment of the present invention;

[0028] Figure 5 A schematic diagram of a process for generating a positive and negative sample set matrix provided by an embodiment of the present invention;

[0029] Figure 6 A schematic diagram of a process for generating a positive and negative sample set matrix provided by an embodiment of the present invention;

[0030] Figure 7 A schematic diagram of the structure of an LSTM model provided by an embodiment of the present invention;

[0031] Figure 8 A schematic diagram of the structure of a neuron state provided by an embodiment of the present invention;

[0032] Fig. 9 A schematic diagram of the structure of a user demand analysis device based on browsing consumption time series data provided by an embodiment of the present invention;

[0033] Fig.10 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0035] In addition, the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be performed in an order different from that shown here.

[0036] Please refer to Figure 1 As shown, the user demand analysis method based on browsing consumption time series data provided by the present invention specifically includes:

[0037] S101 obtains historical behavior data of the user during the monitoring period, and intercepts the behavior data of the user within a predetermined period according to the transaction behavior of the user in the historical behavior data to construct positive sample data and negative sample data;

[0038] S102 performs numerical processing on the positive sample data and the negative sample data to generate a numerical sequence, and performs vector processing on the numerical sequence to generate a positive and negative sample set matrix;

[0039] S103 obtains a demand analysis model by training and verifying a learning model constructed based on a deep learning long short-term memory neural network through the positive and negative sample set matrix;

[0040] S104 collects the user's behavior data to be tested, and obtains the user's needs according to the behavior data to be tested and the needs analysis model.

[0041] In this embodiment, the present invention proposes a method for analyzing user purchase demand based on deep learning, based on the behavioral rules contained in the time series data such as massive user browsing records and historical consumption information, combined with the long short-term memory neural network deep model; among them, deep learning is one of the most concerned technologies in the field of machine learning and artificial intelligence in recent years, and the deep learning model has made major breakthroughs in many fields with its powerful expression ability and network structure. The present invention applies deep learning technology to the user demand analysis business scenario, based on the multi-platform user browsing records and account transaction information time series data, integrates and constructs a long short-term memory neural network deep model, predicts and analyzes the user's demand for products, and realizes the accurate push of products to users, thereby reducing the cost of manpower and material resources, while improving user satisfaction and increasing user retention rate.

[0042] Please refer to Figure 2 As shown, in one embodiment of the present invention, based on the transaction behavior of the user in the historical behavior data, the behavior data of the user within a predetermined period is intercepted to construct positive sample data and negative sample data, which includes:

[0043] S201: when the historical behavior data includes a transaction behavior, first behavior data within a predetermined first period before the transaction occurrence time is intercepted according to the transaction occurrence time of the transaction behavior, and positive sample data is generated according to the first behavior data;

[0044] S202: When the user does not perform any transaction behavior within a predetermined second period in the historical behavior data, intercept the second behavior data of the user within a predetermined third period, and generate negative sample data according to the second behavior data.

[0045] In the above embodiment, when the historical behavior data includes transaction behavior, intercepting the first behavior data within a predetermined first period before the transaction occurrence time according to the transaction occurrence time of the transaction behavior includes: when the historical behavior data includes multiple transaction behaviors, the transaction occurrence time of the Nth transaction behavior is used as the starting time; after moving forward the predetermined fourth period from the starting time, intercepting the first behavior data within the predetermined first period. In another embodiment, when the user does not have a transaction behavior within the predetermined second period in the historical behavior data, intercepting the second behavior data of the user within the predetermined third period includes: the third period is less than the second period, and the start time and end time are different.

[0046] In actual work, user browsing consumption time series data is data that changes over time and can be mainly divided into two categories: 1. Browsing records of users visiting PC and mobile APP; 2. Real-time data of user bank account balance; To this end, by confirming the positive and negative labels and the input data range, you can construct a positive and negative sample set of user browsing consumption time series data. Therefore, in this embodiment, the positive sample label setting rule can be set as follows: During the observation period, users who have purchasing behavior will be set with a positive sample label; if there are multiple purchases, the first purchase will be used as the date of purchase of the product. The positive sample data input is: 15 days before the date the user purchases the product, and then take the previous 30 days of behavior data as the input of the model positive sample, which can be specifically as follows Figure 3 As shown. The reason for setting a 15-day interval is that if the data of the 30 days before the "user purchases the product date" is directly taken, the data of the last day is taken, and there is only a 1-day interval between the purchase behavior. The model will actually become "predict whether the user will buy the product on the next day based on the data of the past 30 days", which is different from the facts; here the goal is to predict the probability of whether the user will demand the product in the next 30 days, and 30 days include the first day of the future and the 30th day of the future, which is an average of 15 days, so the time interval is selected as 15 days. On the other side, the negative sample label setting rule is: during the observation period, users who do not have any purchase records are set with positive sample labels. Among them, the labels of negative samples are better determined. The negative sample data input provided in this embodiment adopts the scheme of taking data from the middle: the middle 30 days of data in the observation period are selected as the input of the negative sample of the model, as shown in the following figure. Figure 4 As shown. The reason why we take data from the middle instead of the end is that we can clearly get the result that the user did not buy the product after the data collection period, so we can confirm that the user is a negative sample; if we only take data from the end, it is not ruled out that some users suddenly buy products after the observation period, so such users should be classified as positive samples; therefore, based on the principle of robustness, we choose to take data from the middle as the input of negative samples of the dynamic model.

[0047] In one embodiment of the present invention, before the positive sample data and the negative sample data are numerically processed to generate a numerical sequence, it also includes: obtaining the URL browsing records in the positive sample data and the negative sample data, eliminating irrelevant URLs according to the correlation between the URL browsing records and the transaction behavior, and obtaining associated URL strings. In this way, the subsequent numerical processing of the positive sample data and the negative sample data to generate a numerical sequence may include: numbering the associated URL strings, constructing an association dictionary of URL strings and URL numbers; generating a numerical sequence according to the association dictionary and the associated URL strings. Based on the above embodiment, in another embodiment of the present invention, vectorizing the numerical sequence to generate a positive and negative sample set matrix includes: mapping the numerical sequence into a multidimensional vector according to a preset unit browsing time to obtain a positive and negative sample set matrix; for details, please refer to Figure 5 shown.

[0048] In actual work, the browsing history of users visiting PC and mobile APP consists of multiple URLs, each of which is a string type. The matrix process of this type of positive and negative sample sets is as follows:

[0049] 1. Data cleaning: remove irrelevant URLs from the browsing records of sample users and retain URL strings related to the products purchased by users.

[0050] 2. Digitalize user browsing records, number all URL strings in user browsing records, and construct a URL string-URL number dictionary. In this way, each user's browsing record can be represented as a numerical sequence consisting of a string of purchased visit URL numbers.

[0051] 3. Vectorize the numerical sequence of user browsing record numbers and map the browsing record number sequence into a multidimensional vector. The dimension of the multidimensional vector is recorded as L. The L value can be set based on the maximum URL sequence length of the actual user browsing record during the observation period and the calculation cost of the LSTM model. If the length of the URL sequence of the user browsing record exceeds L, only the first L URL numbers are retained. If there are less than L, they are filled with 0. In this way, all user browsing records are vectors with L URL numbers. By converting each URL number into an N-dimensional vector, an L×N-dimensional matrix representing each user's access record can be obtained, and each matrix has positive and negative sample labels.

[0052] Please refer to Figure 6As shown, in another embodiment of the present invention, vectorizing the numerical sequence to generate a positive and negative sample set matrix may include: obtaining the user account amount change data in the positive sample data and the negative sample data, and constructing a unit time amount change vector based on the user account amount change data with a preset time as a unit; aggregating the unit time amount change vector to generate a positive and negative sample set matrix. In actual work, the real-time data of the user's bank account amount is numerical data, and the positive and negative sample sets can be directly matrixed. The specific process is as follows:

[0053] 1. Slice the real-time change data of the user's bank account balance on a daily basis to construct a vector of the user's single-day account balance change.

[0054] 2. Aggregate the user's single-day account balance change vectors to construct a 30-day account balance change time series matrix. Each matrix has positive and negative sample labels.

[0055] It is worth noting that the above Figure 5 and Figure 6 The corresponding embodiments thereof can be used in combination in actual work, and relevant technicians in the field can choose to use them in combination according to actual needs, and the present invention is not further limited here.

[0056] In one embodiment of the present invention, obtaining a demand analysis model by training and verifying a learning model constructed based on a deep learning long short-term memory neural network through the positive and negative sample set matrix includes: dividing the positive and negative sample set matrix into a training set and a test set according to a preset ratio; training the learning model through the training set, and verifying the trained learning model through the test set to obtain a demand analysis model. In actual work, the positive and negative sample sets of user browsing consumption time series data can be divided into a training set and a test set, the training set is used to train and construct a user purchase demand prediction model, and the test set is used to test and verify the model prediction effect; this process can be implemented using existing technologies, and will not be described in detail here.

[0057] In one embodiment of the present invention, the learning model mainly adopts the long short-term memory neural network LSTM model in practical application. Of course, relevant technicians in this field can also choose other learning models according to actual needs, and the present invention does not limit this. Among them, the LSTM model provided by the present invention mainly uses a unique gate structure to add or delete information to the neuron state in the neural network, thereby interactively controlling the neuron state, changing the information carried by the neuron state, and protecting and controlling the neuron state; its structure is as follows Figure 7 shown.

[0058] X t-1 , X t , X t+1Represents the user consumption browsing time series data input into the model. t-1、 h t 、h t+1 It represents the hidden state of the recurrent neural network RNN, which can be regarded as the intermediate vector output by each loop node A and saved as state information. The hidden state is both the output of the previous node and passed to the next node as input. The output value h of each loop is t Both with X t and h t-1 Each loop node A of RNN will have an output variable Y for the previous time series variable X. Y at each moment is a function of the current output state h (possibly a Sigmoid or softmax function). LSTM adds a core vector C based on RNN. t , which runs through a horizontal line of the entire structure, is called neuron state information. Figure 8 As shown in:

[0059] The neuron state runs through the entire chain structure, with little linear effect, and mainly plays the role of transmitting information. The transmitted information includes both the information that should be forgotten when passing through each node A and the new input. LSTM proposes the concept of multi-layer gate control, namely the forget gate, input gate and output gate; the core process of the LSTM model is as follows:

[0060] 1. Forget gate: determines the information that should be forgotten by the neuron, which is composed of Sigmoid layers. The forget gate reads the input h of the previous hidden state layer. t-1 and the input X at the current time point t , in neuron state C t-1 Output value f t , f t The value range is [0, 1], "1" means "completely retain information", and "0" means "completely forget information". The specific calculation formula is as follows:

[0061] f t =σ(W f ·[h t-1 , X t ]+b f );

[0062] Among them, W f is the weight matrix of the forget gate, [h t-1 , X t ] means connecting two vectors into a longer vector, b f is the bias term of the forget gate, and σ is the Sigmoid function.

[0063] 2. Input gate: determines the information that should be retained by the neuron. It consists of two parts. The first part determines the value that needs to be updated through the Sigmoid layer, and the second part creates an intermediate value through the tanh layer. Add to the current neuron state C t The specific calculation formula is as follows:

[0064] i t =σ(W i ·[h t-1 , X t ]+b i );

[0065]

[0066] Among them, W i , W C is the weight matrix of the input gate, [h t-1 , X t ] means connecting two vectors into a longer vector, b i 、b C is the bias term of the input gate, σ is the Sigmoid function, and tanh is the tanh function.

[0067] 3. Neuron state update: Combining the forget gate and the input gate can update the neuron state and update the old neuron state C t-1 To the new neuron state C t . Multiply the old neuron state by the forget gate output value f t , forget the information that was previously decided to be forgotten, and then add the product of the two parts of the input gate to update the neuron state to the latest C t Status value; the specific calculation formula is as follows:

[0068]

[0069] 4. Input gate: determines the output information of the neural network, and consists of two parts. The first part uses the Sigmoid layer to determine which part of the neuron state to output, and the second part processes the current state through the tanh layer, and then multiplies it with the output of the first part to ensure that the output is the part that should be output; the specific calculation formula is as follows:

[0070] o t =σ(W o ·[h t-1 , Xt ] +b o );

[0071] h t =o t tanh(C t );

[0072] Among them, W o is the weight matrix of the output gate, [h t-1 , X t ] means connecting two vectors into a longer vector, b o is the bias term of the output gate, σ is the Sigmoid function, and tanh is the tanh function. The final output h t The result calculated from the first part is o t and the neuron state value C t Joint decision.

[0073] After the above four steps, the positive and negative sample sets of user browsing consumption time series data can be effectively trained and predicted. The user purchase demand prediction LSTM model obtained through continuous iterative optimization during the training process can give a prediction score for the user's demand for purchasing products in the next 30 days. The higher the prediction score, the greater the user's demand for the product in the next 30 days and the higher the purchase probability.

[0074] The present invention identifies the user's purchase demand for products in a certain period in the future through three steps: constructing a user's browsing consumption time series positive and negative sample set, matrixing the positive and negative sample set, and constructing a model based on a deep learning long short-term memory neural network, thereby realizing precision marketing to the user; the LSTM model adopted by the present invention can not only process browsing consumption time series data, but also effectively prevent the convergence problem caused by gradient disappearance or explosion, and the prediction effect is also very significant; at the same time, the present invention applies the deep learning long short-term memory neural network technology to the analysis and prediction of user purchase demand, which can not only effectively process time series data, but also automatically obtain data features, and input the quantized time series positive and negative sample sets into the model, and can directly output the user demand analysis and prediction results, which has the advantages of high accuracy, strong generalization ability, and the ability to process massive data.

[0075] Please refer to Fig. 9As shown, the present invention also provides a user demand analysis device based on browsing consumption time series data, the device comprising a sample collection module, a preprocessing module, a model building module and an analysis module; the sample collection module is used to obtain the historical behavior data of the user during the monitoring period, and according to the transaction behavior of the user in the historical behavior data, the behavior data of the user in a predetermined period is respectively intercepted to construct positive sample data and negative sample data; the preprocessing module is used to perform numerical processing on the positive sample data and the negative sample data to generate a numerical sequence, and perform vector processing on the numerical sequence to generate a positive and negative sample set matrix; the model building module is used to train and verify the learning model constructed based on the deep learning long short-term memory neural network through the positive and negative sample set matrix to obtain a demand analysis model; the analysis module is used to collect the user's behavior data to be tested, and obtain user needs according to the behavior data to be tested and the demand analysis model.

[0076] In the above embodiment, the sample collection module includes an extraction unit, which is used to, when the historical behavior data includes transaction behavior, intercept the first behavior data within a predetermined first period before the transaction occurrence time according to the transaction occurrence time of the transaction behavior, and generate positive sample data based on the first behavior data; and, when the user has not performed any transaction behavior within a predetermined second period in the historical behavior data, intercept the second behavior data of the user within a predetermined third period, and generate negative sample data based on the second behavior data.

[0077] In one embodiment of the present invention, the preprocessing module further includes a screening unit, the screening unit is used to obtain the URL browsing records in the positive sample data and the negative sample data, remove irrelevant URLs according to the correlation between the URL browsing records and the transaction behavior, and obtain the associated URL string. Further, in another embodiment of the present invention, the preprocessing module includes a sequence unit and a vector unit; the sequence unit is used to number the associated URL strings, construct an association dictionary of URL strings and URL numbers, and generate a numerical sequence according to the association dictionary and the associated URL strings; the vector unit is used to map the numerical sequence into a multidimensional vector according to the preset unit browsing time to obtain a positive and negative sample set matrix.

[0078] In one embodiment of the present invention, the preprocessing module includes: obtaining user account balance change data in the positive sample data and the negative sample data, constructing a unit time amount change vector based on the user account balance change data with a preset time length as a unit; and aggregating the unit time amount change vector to generate a positive and negative sample set matrix.

[0079] In one embodiment of the present invention, the model construction module includes: dividing the positive and negative sample set matrix into a training set and a test set according to a preset ratio; training the learning model through the training set, and verifying the trained learning model through the test set to obtain a demand analysis model.

[0080] Since the specific implementation methods of each module in the user demand analysis device based on browsing consumption time series data provided by the present invention have been described in detail in the above embodiments, they will not be described in detail one by one here.

[0081] The beneficial technical effect of the present invention is that it accurately analyzes user needs through the behavioral patterns contained in time series data such as massive user browsing records and historical consumption information, combined with a long short-term memory neural network deep model; it can not only process browsing consumption time series data, but also effectively prevent convergence problems caused by gradient disappearance or explosion, and the analysis effect is also very significant.

[0082] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the computer program.

[0083] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for executing the above method.

[0084] like Fig.10 As shown, the electronic device 600 may further include: a communication module 110, an input unit 120, an audio processing unit 130, a display 160, and a power supply 170. It is worth noting that the electronic device 600 does not necessarily have to include Fig.10 In addition, the electronic device 600 may also include Fig.10 For components not shown, reference may be made to the prior art.

[0085] like Fig.10 As shown, the central processor 100 is sometimes also referred to as a controller or an operation control, and may include a microprocessor or other processor devices and / or logic devices. The central processor 100 receives inputs and controls the operations of various components of the electronic device 600.

[0086] The memory 140 may be, for example, one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory or other suitable devices. The above-mentioned information related to the failure may be stored, and a program for executing the relevant information may also be stored. The CPU 100 may execute the program stored in the memory 140 to implement information storage or processing.

[0087] The input unit 120 provides input to the CPU 100. The input unit 120 is, for example, a key or a touch input device. The power supply 170 is used to provide power to the electronic device 600. The display 160 is used to display display objects such as images and text. The display may be, for example, an LCD display, but is not limited thereto.

[0088] The memory 140 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It may also be a memory that saves information even when the power is off, can be selectively erased, and is provided with more data, examples of which are sometimes referred to as EPROMs, etc. The memory 140 may also be some other type of device. The memory 140 includes a buffer memory 141 (sometimes referred to as a buffer). The memory 140 may include an application / function storage unit 142, which is used to store application programs and function programs or processes for executing the operation of the electronic device 600 through the central processor 100.

[0089] The memory 140 may also include a data storage unit 143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 144 of the memory 140 may include various drivers for communication functions of the electronic device and / or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0090] The communication module 110 is a transmitter / receiver 110 that transmits and receives signals via an antenna 111. The communication module (transmitter / receiver) 110 is coupled to the central processor 100 to provide input signals and receive output signals, which may be the same as the case of a conventional mobile communication terminal.

[0091] Based on different communication technologies, multiple communication modules 110 may be provided in the same electronic device, such as a cellular network module, a Bluetooth module and / or a wireless LAN module. The communication module (transmitter / receiver) 110 is also coupled to a speaker 131 and a microphone 132 via an audio processor 130 to provide an audio output via the speaker 131 and receive an audio input from the microphone 132, thereby realizing a common telecommunication function. The audio processor 130 may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 130 is also coupled to the central processor 100, so that the sound can be recorded on the local machine through the microphone 132, and the sound stored on the local machine can be played through the speaker 131.

[0092] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0093] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. 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.

[0094] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate 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 A function specified in one or more boxes.

[0095] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0096] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A user demand analysis method based on browsing consumption time series data, characterized in that: The method comprises: Obtaining historical behavior data of the user during the monitoring period, and intercepting the behavior data of the user within a predetermined period according to the transaction behavior of the user in the historical behavior data to construct positive sample data and negative sample data; Performing numerical processing on the positive sample data and the negative sample data to generate a numerical sequence, and performing vector processing on the numerical sequence to generate a positive and negative sample set matrix; The demand analysis model is obtained by training and verifying the learning model constructed based on the deep learning long short-term memory neural network through the positive and negative sample set matrix; Collecting the user's behavior data to be tested, and obtaining the user's needs according to the behavior data to be tested and the needs analysis model; According to the transaction behavior of the user in the historical behavior data, the behavior data of the user in a predetermined period is intercepted to construct positive sample data and negative sample data, including: When the historical behavior data includes a transaction behavior, first behavior data within a predetermined first period before the transaction occurrence time is intercepted according to the transaction occurrence time of the transaction behavior, and positive sample data is generated according to the first behavior data; When the user does not perform any transaction behavior within a predetermined second period in the historical behavior data, intercepting the user's second behavior data within a predetermined third period, and generating negative sample data according to the second behavior data; Before the positive sample data and the negative sample data are numerically processed to generate a numerical sequence, the method further includes: The website browsing records in the positive sample data and the negative sample data are obtained, and irrelevant websites are eliminated according to the correlation between the website browsing records and the transaction behavior to obtain the associated website character string.

2. The user demand analysis method based on browsing consumption time series data according to claim 1 is characterized in that: When the historical behavior data includes transaction behavior, the first behavior data within a predetermined first period before the transaction occurrence time is intercepted according to the transaction occurrence time of the transaction behavior, including: when the historical behavior data includes multiple transaction behaviors, the transaction occurrence time of the Nth transaction behavior is used as the starting time; after pushing forward a predetermined fourth period with the starting time as the starting point, the first behavior data within the predetermined first period is intercepted.

3. The user demand analysis method based on browsing consumption time series data according to claim 1 is characterized in that: When the user has not performed any transaction within a predetermined second period in the historical behavior data, intercepting the user's second behavior data within a predetermined third period includes: the third period is shorter than the second period, and the start time and the end time are different.

4. The user demand analysis method based on browsing consumption time series data according to claim 1 is characterized in that: The step of performing numerical processing on the positive sample data and the negative sample data to generate a numerical sequence comprises: Numbering the associated URL strings and constructing an associated dictionary of URL strings and URL numbers; A numerical sequence is generated according to the associated dictionary and the associated URL character string.

5. The user demand analysis method based on browsing consumption time series data according to claim 4 is characterized in that: Performing vectorization processing on the numerical sequence to generate a positive and negative sample set matrix includes: mapping the numerical sequence into a multi-dimensional vector according to a preset unit browsing time to obtain a positive and negative sample set matrix.

6. The user demand analysis method based on browsing consumption time series data according to claim 1 is characterized in that: Vectorizing the numerical sequence to generate a positive and negative sample set matrix includes: Acquire the user account balance change data in the positive sample data and the negative sample data, and construct a unit time amount change vector based on the user account balance change data with a preset time length as a unit; The unit time amount change vectors are aggregated to generate a positive and negative sample set matrix.

7. The user demand analysis method based on browsing consumption time series data according to claim 1 is characterized in that: The demand analysis model is obtained by training and verifying the learning model constructed based on the deep learning long short-term memory neural network through the positive and negative sample set matrix, including: Dividing the positive and negative sample set matrix into a training set and a test set according to a preset ratio; The learning model is trained by the training set, and the demand analysis model is obtained by validating the trained learning model by the test set.

8. A user demand analysis device based on browsing consumption time series data, characterized in that: The device comprises a sample collection module, a preprocessing module, a model building module and an analysis module; The sample collection module is used to obtain the historical behavior data of the user during the monitoring period, and to intercept the behavior data of the user within a predetermined period according to the transaction behavior of the user in the historical behavior data to construct positive sample data and negative sample data; The preprocessing module is used to perform numerical processing on the positive sample data and the negative sample data to generate a numerical sequence, and perform vector processing on the numerical sequence to generate a positive and negative sample set matrix; The model building module is used to train and verify the learning model built based on the deep learning long short-term memory neural network through the positive and negative sample set matrix to obtain a demand analysis model; The analysis module is used to collect the user's behavior data to be tested, and obtain the user's needs according to the behavior data to be tested and the needs analysis model; The sample collection module includes an extraction unit, which is used to intercept first behavior data within a predetermined first period before the transaction occurrence time according to the transaction occurrence time of the transaction behavior when the historical behavior data contains transaction behavior, and generate positive sample data according to the first behavior data; and when the user does not have a transaction behavior within a predetermined second period in the historical behavior data, intercept the user's second behavior data within a predetermined third period, and generate negative sample data according to the second behavior data; The preprocessing module further includes a screening unit, which is used to obtain website browsing records in the positive sample data and the negative sample data, eliminate irrelevant websites according to the correlation between the website browsing records and the transaction behavior, and obtain associated website character strings.

9. The user demand analysis device based on browsing consumption time series data according to claim 8, characterized in that: The preprocessing module includes a sequence unit and a vector unit; The sequence unit is used to number the associated website string, construct an association dictionary of website string and website number, and generate a numerical sequence according to the association dictionary and the associated website string; The vector unit is used to map the numerical sequence into a multi-dimensional vector to obtain a positive and negative sample set matrix according to a preset unit browsing time.

10. The user demand analysis device based on browsing consumption time series data according to claim 8, characterized in that: The preprocessing module includes: obtaining user account balance change data in the positive sample data and the negative sample data, constructing a unit time amount change vector based on the user account balance change data with a preset time as a unit; and aggregating the unit time amount change vector to generate a positive and negative sample set matrix.

11. The user demand analysis device based on browsing consumption time series data according to claim 8, characterized in that: The model building module includes: dividing the positive and negative sample set matrix into a training set and a test set according to a preset ratio; training the learning model through the training set, and verifying the trained learning model through the test set to obtain a demand analysis model.

12. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program for executing the method according to any one of claims 1 to 7.

Images