A device access loss early warning method and device based on dynamic data

Abstract

The present invention discloses a device access loss early warning method and device based on dynamic data, the method comprising: respectively obtaining device cross-sectional data and device panel data refreshed over time when the device to be identified uses a product; inputting the device cross-sectional data into a memory model in a device access loss model to obtain first feature data; inputting the device panel data into a generalization model in a device access loss model to obtain second feature data; inputting the first feature data and the second feature data into an output layer in a device access loss model to obtain a device access loss probability; determining a device access loss level according to the device access loss probability, and issuing a loss early warning. The device access loss probability determined by the present invention can reflect the loss of user devices during the product use cycle, thereby effectively avoiding the loss of user devices during the product use cycle, and further ensuring corporate profits and property safety during the entire product use cycle.

Description

Technical Field

[0001] The present invention relates to the field of device data processing, and in particular to a method, device, electronic device and computer-readable medium for early warning of device access loss based on dynamic data. Background Art

[0002] With the continuous development of society, the competition in the industry is becoming more and more fierce. Research shows that under the current market conditions, retaining old user devices can bring greater economic benefits than tapping new user devices. Therefore, attention to old user devices and prevention of loss have become the focus of attention.

[0003] However, in the task of predicting user device access loss using traditional tree models, user loss is generally predicted through device feature data at a certain moment, while some device feature data is constantly refreshed and accumulated during the use of the product. If only the device feature data at a certain moment is used to predict the user device access loss, the prediction result is only applicable to a specific moment (such as when using the product), and cannot accurately reflect the loss of user devices during the product use cycle. This is especially true for rental products, such as book borrowing, shared bicycles, etc. The loss of user device access during the product use cycle will not only reduce corporate profits, but may also cause the product to be unable to be returned, causing economic losses to the company. Therefore, it is urgent to find a user churn warning method that can reflect the loss of user devices during the product use cycle to ensure the property and data security of the company. Summary of the invention

[0004] In view of this, the main purpose of the present invention is to propose a device access loss warning method, device, electronic device and computer-readable medium based on dynamic data, in order to at least partially solve at least one of the above-mentioned technical problems.

[0005] In order to solve the above technical problems, the first aspect of the present invention proposes a device access loss warning method based on dynamic data, the method comprising:

[0006] Respectively obtain the equipment cross-section data when the equipment to be identified uses the product and the equipment panel data updated over time;

[0007] Inputting the device cross-section data into a memory model of a device access loss model for processing, and obtaining first feature data that affects device access loss when the device to be identified uses the product;

[0008] Inputting the device panel data into a generalized model of a device access loss model for processing, and obtaining second feature data that is updated over time and has an impact on device access loss;

[0009] Inputting the first feature data and the second feature data into the output layer of the device access loss model for full connection processing to obtain the device access loss probability;

[0010] A device access loss level is determined according to the device access loss probability, and a device access loss warning is issued according to the device access loss level.

[0011] According to a preferred embodiment of the present invention, the generalized model includes: an input layer and a hidden layer connected to the input layer, wherein: the input layer includes a plurality of input nodes, each input node corresponds to a hidden node connected to the hidden layer; the device panel data is input into the generalized model of the device access loss model for processing, and the second feature data that is refreshed over time and has an impact on the device access loss is obtained, including:

[0012] Arranging the device panel data in chronological order to obtain a device panel data sequence;

[0013] Input the device panel data sequence into each input node in sequence;

[0014] For each hidden node, the output data of the previous hidden node and the device panel data of the corresponding input node are processed to obtain output data;

[0015] The output data of the last hidden node is used as the second feature data.

[0016] According to a preferred embodiment of the present invention, the output data h(t) of the current hidden node is:

[0017] h(t)=f(W*X(t)+U*h(t-1)+b);

[0018] Wherein: h(t-1) is the output data of the previous hidden node, X(t) is the device panel data of the input node corresponding to the current hidden node; f is the activation function, W is the first shared parameter, U is the second shared parameter, and b is the third shared parameter.

[0019] According to a preferred embodiment of the present invention, the number of input nodes and hidden nodes in the generalization model is variable, and the hidden layer uses the output data of the last device panel data in the device panel data sequence as the second feature data.

[0020] According to a preferred embodiment of the present invention, the output layer is fully connected by the following formula:

[0021] y = σ(W1X1+W2X2+c);

[0022] Wherein: y is the probability of device access loss, X1 is the first feature data, X2 is the second feature data, W1 is the parameter corresponding to the first feature data, W2 is the parameter corresponding to the second feature data; σ is the sigmoid activation function, and c is the bias term.

[0023] According to a preferred embodiment of the present invention, the device access loss model is pre-trained in the following manner:

[0024] Obtain equipment cross-section data when historical equipment uses the product, historical equipment panel data refreshed over time, and historical equipment tag data;

[0025] Initialize various parameters, input the device cross-section data of historical devices into the memory model of the device access loss model, and input the device panel data of historical devices into the generalized model of the device access loss model;

[0026] Calculate the loss function based on the output of the device access loss model and the historical device label data;

[0027] Perform reverse calculation based on the loss function to update the parameters.

[0028] In order to solve the above technical problems, the second aspect of the present invention provides a device access loss warning device based on dynamic data, the device comprising:

[0029] An acquisition module is used to respectively acquire the device cross-sectional data of the device to be identified when the product is used and the device panel data updated over time;

[0030] A first input module, used for inputting the device cross-section data into a memory model in a device access loss model for processing, and obtaining first feature data that affects device access loss when the device to be identified uses the product;

[0031] A second input module is used to input the device panel data into a generalized model in the device access loss model for processing, so as to obtain second feature data that affects the device access loss and is updated over time;

[0032] A third input module, configured to input the first feature data and the second feature data into an output layer of a device access loss model for full connection processing to obtain a device access loss probability;

[0033] The early warning module is used to determine the device access loss level according to the device access loss probability and issue a device access loss early warning.

[0034] According to a preferred embodiment of the present invention, the generalization model comprises: an input layer and a hidden layer connected to the input layer, wherein: the input layer comprises a plurality of input nodes, each input node corresponds to a hidden node connected to the hidden layer, and the second input module comprises:

[0035] A sorting module, used for arranging the device panel data in chronological order to obtain a device panel data sequence;

[0036] A sub-input module, used for inputting the device panel data sequence into each input node in sequence;

[0037] A processing module, used for processing the output data of each hidden node on the previous hidden node and the device panel data of the corresponding input node to obtain output data;

[0038] The output module is used to use the output data of the last hidden node as the second feature data.

[0039] According to a preferred embodiment of the present invention, the processing module obtains the output data h(t) of the current hidden node through the following formula:

[0040] h(t)=f(W*X(t)+U*h(t-1)+b);

[0041] Wherein: h(t-1) is the output data of the previous hidden node, X(t) is the device panel data of the input node corresponding to the current hidden node; f is the activation function, W is the first shared parameter, U is the second shared parameter, and b is the third shared parameter.

[0042] According to a preferred embodiment of the present invention, the number of input nodes and hidden nodes in the generalization model is variable, and the output module uses the output data of the last device panel data in the device panel data sequence as the second feature data.

[0043] According to a preferred embodiment of the present invention, the output layer is fully connected by the following formula:

[0044] y = σ(W1X1+W2X2+c);

[0045] Wherein: y is the probability of device access loss, X1 is the first feature data, X2 is the second feature data, W1 is the parameter corresponding to the first feature data, W2 is the parameter corresponding to the second feature data; σ is the sigmoid activation function, and c is the bias term.

[0046] According to a preferred embodiment of the present invention, the device further comprises:

[0047] A pre-acquisition module is used to respectively acquire the equipment cross-section data of the historical equipment to be identified when the product is used, the historical equipment panel data refreshed over time, and the historical equipment label data;

[0048] A pre-input module, used to initialize various parameters, input the device cross-section data of historical devices into the memory model of the device access loss model, and input the device panel data of historical devices into the generalized model of the device access loss model;

[0049] A calculation module, used to calculate a loss function based on the output result of the device access loss model and historical device label data;

[0050] The updating module is used to perform reverse operation based on the loss function to update parameters.

[0051] In order to solve the above technical problems, the third aspect of the present invention provides an electronic device, including:

[0052] Processor; and

[0053] A memory storing computer executable instructions, which when executed cause the processor to perform the above method.

[0054] To solve the above technical problem, the fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores one or more programs, and when the one or more programs are executed by a processor, the above method is implemented.

[0055] The present invention takes into account that user device access loss is a dynamic process, and obtains device cross-sectional data when the device to be identified uses a product and device panel data refreshed over time respectively; numerically processes the device cross-sectional data through a memory model in a device access loss model to obtain first feature data that affects device access loss when the device to be identified uses the product; encodes the device panel data through a generalized model in the device access loss model to obtain second feature data that affects device access loss refreshed over time, and then fully connects the first feature data and the second feature data through an output layer in the device access loss model. The device access loss probability determined in this way can reflect the loss of user devices during the product use cycle, and then issues a device access loss warning after determining the device access loss level based on the device access loss probability, thereby effectively avoiding the loss of equipment during the product use cycle and ensuring corporate profits and property safety during the entire product use cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to make the technical problems solved by the present invention, the technical means adopted and the technical effects achieved more clearly, the specific embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, it should be noted that the drawings described below are only drawings of exemplary embodiments of the present invention, and those skilled in the art can obtain drawings of other embodiments based on these drawings without creative work.

[0057] Figure 1 It is a flow chart of a device access loss warning method based on dynamic data according to an embodiment of the present invention;

[0058] Figure 2 It is a schematic diagram of the structural framework of the device access loss model according to an embodiment of the present invention;

[0059] Figure 3 It is a schematic diagram of the structural framework of the generalized model of an embodiment of the present invention;

[0060] Figure 4 It is a schematic diagram of the structural framework of a device access loss warning device based on dynamic data according to an embodiment of the present invention;

[0061] Figure 5 is a structural block diagram of an exemplary embodiment of an electronic device according to the present invention;

[0062] Figure 6 is a schematic diagram of an embodiment of a computer readable medium of the present invention. DETAILED DESCRIPTION

[0063] The exemplary embodiments of the present invention will now be described more fully with reference to the accompanying drawings. Although each exemplary embodiment can be implemented in a variety of specific ways, it should not be understood that the present invention is limited to the embodiments set forth herein. On the contrary, these exemplary embodiments are provided to make the content of the present invention more complete and to facilitate the comprehensive communication of the inventive concept to those skilled in the art.

[0064] The same reference numerals in the drawings represent the same or similar elements, components or parts, and thus the repeated description of the same or similar elements, components or parts may be omitted below. It should also be understood that although the first, second, third and other attributives representing the numbers may be used herein to describe various devices, elements, components or parts, these devices, elements, components or parts should not be limited by these attributives. In other words, these attributives are only used to distinguish one from another. For example, the first device may also be called the second device, but it does not deviate from the essential technical solution of the present invention. In addition, the terms "and / or" and "and / or" refer to all combinations including any one or more of the listed items.

[0065] See also Figure 1 , Figure 1The present invention provides a device access loss early warning method based on dynamic data, such as Figure 1 As shown, the method includes:

[0066] S1. Obtain the device cross-sectional data of the device to be identified when using the product and the device panel data updated over time;

[0067] The present invention takes into account that user device access loss is a dynamic process, and comprehensively describes the user device access loss process with device cross-sectional data and device panel data, wherein: device cross-sectional data refers to device data that can be obtained when the user device uses the product, also known as static device data. Panel device data refers to device data that can be continuously refreshed over time, also known as dynamic device data.

[0068] Among them: device data can be data disclosed by users of the device on the Internet service platform. The data may include one or more of the following: user name, user age, user occupation, user income, user place of origin, user authorization information on the Internet platform, user product application behavior information, user security credibility and other public information, but not limited to this. The data processing of this scheme can also be performed only through user information that cannot identify the user's identity, such as age, education, household registration, etc., to protect user privacy; the protection of user privacy can be achieved by deleting or anonymizing the information that can identify the user in the user information, and the anonymization process can be to process the data through encryption means.

[0069] Considering that the device authorization information and the device application product behavior information are generated when the user device uses the product and have a great impact on the device access loss, in this embodiment, the device authorization information on the Internet platform and the device application product behavior information are selected as the device cross-sectional data. Among them: the device authorization information on the Internet platform refers to the number of products authorized by the Internet platform to the device based on the device's historical security credibility and / or the device's product usage record. The device application product behavior information may include: the number of device application products, the return period of the device application product, etc. In this embodiment, the product form can be diverse, such as: physical objects, services, data information, technology, etc. Considering that the device security credibility will be refreshed every time the device uses the product and has a great impact on the device access loss, in this embodiment, the device security credibility is selected as the device cross-sectional data. Among them: the device security credibility can be the evaluation result of the device credit by the Internet platform or a third party based on the device's historical product usage behavior.

[0070] Exemplarily, in this step, when the device uses the product, the authorization information of the device on the Internet platform and the behavior information of the device applying for the product are obtained as the device cross-sectional data, where: the authorization information of the device on the Internet platform is only obtained when the device uses the product, and the behavior information of the device applying for the product can be obtained within a predetermined time window, for example: the predetermined time window can be set from the time when the device is successfully authorized to the time when the device uses the product this time, then between the time when the device is successfully authorized to the time when the device uses the product this time, the number of products applied for by the device, the number of return periods for the products applied for by the device, and other information are obtained. In addition, when the device uses the product, the device security credibility of the device when the device has used the product in previous times is obtained as the device panel data.

[0071] S2, inputting the device cross-section data into a memory model of a device access loss model for processing, and obtaining first feature data that affects device access loss when the device to be identified uses the product;

[0072] S3, inputting the device panel data into a generalized model of a device access loss model for processing, and obtaining second feature data that is updated over time and has an impact on device access loss;

[0073] S4, inputting the first feature data and the second feature data into the output layer of the device access loss model for full connection processing to obtain the device access loss probability;

[0074] In this embodiment, Figure 2 As shown, the device access loss model includes: a memory model, which is used to process device cross-sectional data to obtain first feature data that affects device access loss when the device uses a product; a generalization model, which is used to process device panel data to obtain second feature data that is refreshed over time and affects device access loss; and an output layer connected to the memory model and the generalization model respectively, and the output layer performs full connection processing on the first feature data and the second feature data to obtain the device access loss probability.

[0075] Among them: the memory model is used to learn the first feature data that affects the device access loss when the device uses the product. The memory model can adopt a traditional machine learning model, such as a linear model. Since the device cross-sectional data can be: continuous feature data, discrete feature data, and categorical feature data, in this embodiment, in order to capture the interactive information between the device cross-sectional data input into the memory model and increase the dimension of the device cross-sectional data, the discrete feature data in the device cross-sectional data can be combined to obtain cross-features. For example: discrete features: age and occupation, then after combining age and occupation, the cross-feature age-occupation can be obtained. The categorical feature data in the device cross-sectional data can be numerically processed through one-hot encoding or sequence encoding, and the processed data is input into the memory model to obtain the first feature data that affects the device access loss when the device uses the product.

[0076] The generalization model is used to learn the second feature data that is updated over time and affects the device access loss. Figure 3 , the generalization model includes: an input layer and a hidden layer connected to the input layer, wherein: the input layer may have multiple input nodes X, each input node X may input device panel data at different refresh times, the hidden layer may also have multiple hidden nodes h, each hidden node h may correspond to an input node X, then each hidden node h encodes the output data of the previous hidden node h and the device panel data of the corresponding input node X, that is, the input data of the current hidden node h(t) is the output data of the previous hidden node h(t-1) and the input data of the corresponding input node X(t), through this cycle, the data is added in the hidden layer. The generalization model may adopt a deep neural network or a deep convolutional neural network, which has a multi-layer superimposed linear and nonlinear structure. The deep neural network can capture key features directly from the original data, so it is very suitable for learning features contained in unstructured data. Correspondingly, the device panel data is input into the generalization model of the device access loss model for encoding processing, and the second feature data that affects the device access loss that is refreshed over time may include:

[0077] S31, arranging the device panel data in chronological order to obtain a device panel data sequence;

[0078] For example, if the device panel data is the device safety credibility of the device when the device is used in the past, the device safety credibility of the device when the device is used in the past is sorted according to the time sequence of the device using the product, and the device panel data sequence {X (t-n+1) ,X (t-n+2) …X (t-i) …X (t-1) ,X (t)}; where: n is the total number of times the device uses the product, t is the time when the device uses the product this time (i.e., the nth time), ti is the time when the device uses the product for the (nith) time, X (t-1) The device security credibility when the device uses the product for the first time.

[0079] In addition, in order to facilitate the processing of the generalized model, the panel data of each device can be vectorized first to obtain the panel device vector, and then the panel device vectors can be sorted according to the time series to obtain the device panel data sequence.

[0080] S32, inputting the device panel data sequence into each input node in sequence;

[0081] S33, for each hidden node, processing the output data of the previous hidden node and the device panel data of the corresponding input node to obtain output data;

[0082] For example, Figure 3 , the output data h(t) of the current hidden node can be expressed as:

[0083] h(t)=f(W*X(t)+U*h(t-1)+b);

[0084] Among them: h(t-1) is the output data of the previous hidden node, X(t) is the device panel data of the input node corresponding to the current hidden node; f is the activation function, W is the first shared parameter, U is the second shared parameter, b is the third shared parameter, and the initial values ​​of the first shared parameter, the second shared parameter and the third shared parameter are preset constants, which can be set by the user. During the iterative training process of the model, the first step of calculation is completed with the preset constants, and then the parameters are continuously iteratively updated through the back-propagation composite function derivative. When the model training is completed, the values ​​of the first shared parameter, the second shared parameter and the third shared parameter are determined.

[0085] S34. Use the output data of the last hidden node as the second feature data.

[0086] In this embodiment, the device refreshes and obtains the device panel data each time the product is used. Since each device is at a different point in the product use cycle, the refresh times of the device panel data of each device are also different. Therefore, the number of input nodes and hidden nodes in the generalized model can be variable, so that different numbers of device panel data of each device can be processed. At this time, the output data of the last device panel data in the device panel data sequence is used as the second feature data. For example: before each hidden node processes data, first determine whether the device panel data input by the corresponding input node is the last device panel data in the device panel data sequence. If so, the output data of the previous hidden node and the device panel data of the corresponding input node are processed to obtain output data, and the output data is output to the output layer as the second feature data. If not, the output data of the previous hidden node and the device panel data of the corresponding input node are processed to obtain output data, which is input to the next hidden node.

[0087] Wherein: the input end of the output layer is connected to the output end of the memory model and the output end of the hidden node of the outermost layer of the generalization model respectively, and the first feature data and the second feature data are fully connected. Exemplarily, the output layer can be activated using a sigmoid function to compress the value of the first feature data and the second feature data after full connection to between [0, 1]. Wherein: the output layer is fully connected using the following formula:

[0088] y = σ(W1X1+W2X2+c);

[0089] Wherein: y is the probability of device access loss, X1 is the first feature data, X2 is the second feature data, W1 is the parameter corresponding to the first feature data, W2 is the parameter corresponding to the second feature data; σ is the sigmoid activation function, and c is the bias term.

[0090] In actual application, when a device uses a product, the authorization information of the device on the Internet platform and the device product application behavior information are obtained and input into the memory model. At the same time, the device security credibility of the device during previous uses of the product is obtained and combined into a vector sequence to input into the generalization model. By adjusting the hyperparameters of the device access loss model (such as hidden layer dimensions, etc.), the model is further optimized to obtain a dynamic device access loss probability that can reflect the loss of the device during the product usage cycle. In this way, each time the device uses the product, its loss N days after use can be predicted.

[0091] S5. Determine a device access loss level according to the device access loss probability, and issue a device access loss warning according to the device access loss level.

[0092] After obtaining the probability of device access loss, this embodiment can determine the level of device access loss based on the probability of device access loss, and then issue a device access loss warning according to the loss level of the device. Among them: the device access loss warning can be retention information pushed to the device, for example: the device access loss level can be determined according to the interval where the device access loss probability is located, and the corresponding retention information can be pushed to the device according to the device access loss level. The retention information can be: product discount information, Internet platform red envelopes, etc. In addition, when the device access loss level is lower than the threshold, it is determined that the device may not return the product within the product usage cycle. The device access loss warning can be return information pushed to the device, prompting the device to return the product on time; it can also be a blacklist sent to Internet platform staff so that the staff can focus on monitoring the device to avoid economic losses.

[0093] In this embodiment, further, the device access loss probability that changes with time is determined, and the device access loss level is determined according to the change trend of the device access loss probability that changes with time, and a device access loss warning is issued according to the device access loss level.

[0094] Specifically, the device access churn probability curve can be fitted according to the device access churn probability that changes with time, and the churn probability growth rate can be obtained through the device access churn probability curve; finally, the device access churn probability at the current time point and the churn probability growth rate at the current time point are comprehensively considered to determine the device access churn level to be identified.

[0095] Specifically, the churn grade score can be calculated by the device access churn probability at the current time point and the churn probability growth rate at the current time point, and the churn grade can be determined according to the churn grade score. The churn grade score can be calculated by the following formula:

[0096] F=A a ×a+B×b / b 大 ;

[0097] Where F is the churn level, a is the device access churn probability, b is the churn probability growth rate, b 大 can be a constant set according to the actual situation, or it can be the maximum loss probability growth rate per unit time; A is a constant greater than 1, and B is a constant greater than 0. In this formula, as a increases, A a The greater the probability of device access loss, the greater the impact on the loss grade score. The growth rate of the device access loss probability is normalized to be equal to the magnitude of the device access loss probability, ensuring that the final calculated loss grade score accurately identifies the device to be identified.

[0098] Furthermore, the present invention may pre-train the device access loss model, and the training process may include:

[0099] S11, respectively obtaining equipment cross-sectional data when historical equipment uses the product, equipment panel data of historical equipment refreshed over time, and historical equipment label data;

[0100] The device tag data is used to reflect the actual loss of the device, which may include loss and no loss. For example, 0 represents loss and 1 represents no loss.

[0101] S12, initializing various parameters, inputting the device cross-section data of historical devices into the memory model of the device access loss model, and inputting the device panel data of historical devices into the generalized model of the device access loss model;

[0102] S13, calculating a loss function according to the output result of the device access loss model and the historical device label data;

[0103] Among them: the loss function L is:

[0104]

[0105] in: is the label value of the i-th historical device, y (i) is the predicted value of the access loss probability of the i-th historical device, and N is the total number of historical devices.

[0106] S14. Perform reverse operation based on the loss function to update parameters.

[0107] In this embodiment, matrix linear operations are used between the hidden nodes of the generalized model and between the outermost hidden nodes and the output layer, so the inverse operation can be used to implement parameter update.

[0108] In the reverse operation process, the total model error is first calculated according to the loss function, the error is reversed, and the various parameters in the model are updated.

[0109] Then, according to the updated parameters, steps S12 to S14 are repeated, and the iteration is continued until the model error is within a preset range, and the iteration is stopped to complete the training of the device access loss model.

[0110] Figure 4 The present invention is a device access loss early warning device, such as Figure 4 As shown, the device comprises:

[0111] An acquisition module 41 is used to respectively acquire the device cross-sectional data of the device to be identified when the product is used and the device panel data updated over time;

[0112] A first input module 42 is used to input the device cross-section data into a memory model in a device access loss model for processing, so as to obtain first feature data that affects device access loss when the device to be identified uses the product;

[0113] A second input module 43 is used to input the device panel data into a generalized model in the device access loss model for processing, so as to obtain second feature data that is updated over time and has an impact on the device access loss;

[0114] A third input module 44 is used to input the first feature data and the second feature data into the output layer of the device access loss model for full connection processing to obtain a device access loss probability;

[0115] The early warning module 45 is used to determine the device access loss level according to the device access loss probability and issue a device access loss early warning.

[0116] In a specific implementation, the generalization model includes: an input layer and a hidden layer connected to the input layer, wherein: the input layer includes a plurality of input nodes, each input node corresponds to a hidden node connected to the hidden layer, and the second input module 43 includes:

[0117] A sorting module, used for arranging the device panel data in chronological order to obtain a device panel data sequence;

[0118] A sub-input module, used for inputting the device panel data sequence into each input node in sequence;

[0119] A processing module, used for processing the output data of each hidden node on the previous hidden node and the device panel data of the corresponding input node to obtain output data;

[0120] The output module is used to use the output data of the last hidden node as the second feature data.

[0121] Optionally, the processing module obtains the output data h(t) of the current hidden node by the following formula:

[0122] h(t)=f(W*X(t)+U*h(t-1)+b);

[0123] Wherein: h(t-1) is the output data of the previous hidden node, X(t) is the device panel data of the input node corresponding to the current hidden node; f is the activation function, W is the first shared parameter, U is the second shared parameter, and b is the third shared parameter.

[0124] Optionally, the number of input nodes and hidden nodes in the generalization model is variable, and the output module uses the output data of the last device panel data in the device panel data sequence as the second feature data.

[0125] In a specific implementation, the output layer is fully connected using the following formula:

[0126] y = σ(W1X1+W2X2+c);

[0127] Wherein: y is the probability of device access loss, X1 is the first feature data, X2 is the second feature data, W1 is the parameter corresponding to the first feature data, W2 is the parameter corresponding to the second feature data; σ is the sigmoid activation function, and c is the bias term.

[0128] Furthermore, the device also includes:

[0129] A pre-acquisition module is used to respectively acquire the equipment cross-section data of the historical equipment to be identified when the product is used, the historical equipment panel data refreshed over time, and the historical equipment label data;

[0130] A pre-input module, used to initialize various parameters, input the device cross-section data of historical devices into the memory model of the device access loss model, and input the device panel data of historical devices into the generalized model of the device access loss model;

[0131] A calculation module, used to calculate a loss function based on the output result of the device access loss model and historical device label data;

[0132] The updating module is used to perform reverse operation based on the loss function to update parameters.

[0133] Those skilled in the art will appreciate that the modules in the above device embodiments may be distributed in the device as described, or may be changed accordingly and distributed in one or more devices different from the above embodiments. The modules in the above embodiments may be combined into one module, or may be further split into multiple submodules.

[0134] The electronic device embodiment of the present invention is described below, and the electronic device can be regarded as a physical implementation of the method and device embodiments of the present invention described above. The details described in the electronic device embodiment of the present invention should be regarded as a supplement to the above method or device embodiments; details not disclosed in the electronic device embodiment of the present invention can be implemented with reference to the above method or device embodiments.

[0135] Figure 5 is a structural block diagram of an exemplary embodiment of an electronic device according to the present invention. Figure 5The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0136] like Figure 5 As shown, the electronic device 500 of this exemplary embodiment is in the form of a general data processing device. The components of the electronic device 500 may include, but are not limited to: at least one processing unit 510, at least one storage unit 520, a bus 530 connecting different electronic device components (including the storage unit 520 and the processing unit 510), a display unit 540, etc.

[0137] The storage unit 520 stores a computer-readable program, which may be a source program or a code of a read-only program. The program may be executed by the processing unit 510, so that the processing unit 510 performs the steps of various embodiments of the present invention. For example, the processing unit 510 may perform the following steps: Figure 1 Steps shown.

[0138] The storage unit 520 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 5201 and / or a cache storage unit 5202, and may further include a read-only storage unit (ROM) 5203. The storage unit 520 may also include a program / utility 5204 having a set (at least one) of program modules 5205, such program modules 5205 include but are not limited to: operating electronic devices, one or more application programs, other program modules and program data, each of which or some combination may include the implementation of a network environment.

[0139] Bus 530 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0140] The electronic device 500 may also communicate with one or more external devices 100 (e.g., keyboard, display, network device, Bluetooth device, etc.) so that a user can interact with the electronic device 500 via these external devices 100, and / or the electronic device 500 can communicate with one or more other data processing devices (e.g., routers, modems, etc.). Such communication may be performed through an input / output (I / O) interface 550, and may also be performed through a network adapter 560 with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet). The network adapter 560 may communicate with other modules of the electronic device 500 via the bus 530. It should be understood that although Figure 5Not shown, other hardware and / or software modules may be used in the electronic device 500, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID electronic devices, tape drives, and data backup storage electronic devices.

[0141] Figure 6 Schematic diagram of a computer readable medium embodiment of the present invention. Figure 6 As shown, the computer program can be stored on one or more computer-readable media. The computer-readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electronic device, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. When the computer program is executed by one or more data processing devices, the computer-readable medium can implement the above-mentioned method of the present invention, namely: respectively obtaining device cross-sectional data when the device to be identified uses the product and device panel data refreshed over time; inputting the device cross-sectional data into the memory model of the device access loss model for processing, and obtaining first feature data that affects device access loss when the device to be identified uses the product; inputting the device panel data into the generalization model of the device access loss model for processing, and obtaining second feature data that affects device access loss refreshed over time; inputting the first feature data and the second feature data into the output layer of the device access loss model for full connection processing, and obtaining the device access loss probability; determining the device access loss level according to the device access loss probability, and issuing a device access loss warning according to the device access loss level.

[0142] The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the present invention is not inherently related to any specific computer, virtual device or electronic device, and various general devices can also implement the present invention. The above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A device access loss early warning method based on dynamic data, characterized in that: The method comprises: Obtain the device cross-sectional data of the device to be identified when using the product and the device panel data refreshed over time; the device cross-sectional data refers to the device data that can be obtained when the user device uses the product, also known as static device data; the panel device data refers to the device data that can be continuously refreshed over time, also known as dynamic device data; Inputting the device cross-section data into a memory model of a device access loss model for processing, and obtaining first feature data that affects device access loss when the device to be identified uses the product; Inputting the device panel data into a generalized model of a device access loss model for processing, and obtaining second feature data that is updated over time and has an impact on device access loss; Inputting the first feature data and the second feature data into the output layer of the device access loss model for full connection processing to obtain the device access loss probability; Determine a device access loss level according to the device access loss probability, and issue a device access loss warning according to the device access loss level; Wherein: the generalized model includes: an input layer and a hidden layer connected to the input layer, wherein: the input layer includes a plurality of input nodes, each input node corresponds to a hidden node connected to the hidden layer; the device panel data is input into the generalized model of the device access loss model for processing, and the second feature data that is updated over time and has an impact on the device access loss is obtained, including: Arranging the device panel data in chronological order to obtain a device panel data sequence; Input the device panel data sequence into each input node in sequence; For each hidden node, the output data of the previous hidden node and the device panel data of the corresponding input node are processed to obtain the output data; specifically, the output data h(t) of the current hidden node is: ； Where: h(t-1) is the output data of the previous hidden node, X(t) is the device panel data of the input node corresponding to the current hidden node; f is the activation function, W is the first shared parameter, U is the second shared parameter, and b is the third shared parameter; The output data of the last hidden node is used as the second feature data; The number of input nodes and hidden nodes in the generalized model is variable, and the hidden layer uses the output data of the last device panel data in the device panel data sequence as the second feature data; The memory model is a machine learning model.

2. The method according to claim 1, characterized in that The output layer is fully connected using the following formula: ； Wherein: y is the probability of device access loss, X1 is the first feature data, X2 is the second feature data, W1 is the parameter corresponding to the first feature data, W2 is the parameter corresponding to the second feature data; σ is the sigmoid activation function, and c is the bias term.

3. The method according to claim 1, characterized in that The device access churn model is pre-trained in the following manner: Obtain equipment cross-section data when historical equipment uses the product, historical equipment panel data refreshed over time, and historical equipment tag data; Initialize various parameters, input the device cross-section data of historical devices into the memory model of the device access loss model, and input the device panel data of historical devices into the generalized model of the device access loss model; Calculate the loss function based on the output of the device access loss model and the historical device label data; Perform reverse calculation based on the loss function to update the parameters.

4. A device access loss warning device based on dynamic data, characterized in that: The device comprises: The acquisition module is used to respectively acquire the device cross-sectional data of the device to be identified when using the product and the device panel data refreshed over time; the device cross-sectional data refers to the device data that can be acquired when the user device uses the product, also known as static device data; the panel device data refers to the device data that can be continuously refreshed over time, also known as dynamic device data; A first input module, used for inputting the device cross-section data into a memory model in a device access loss model for processing, and obtaining first feature data that affects device access loss when the device to be identified uses the product; A second input module is used to input the device panel data into a generalized model in the device access loss model for processing, so as to obtain second feature data that affects the device access loss and is updated over time; A third input module, configured to input the first feature data and the second feature data into an output layer of a device access loss model for full connection processing to obtain a device access loss probability; An early warning module, used to determine the device access loss level according to the device access loss probability, and issue a device access loss early warning; Wherein: the generalization model includes: an input layer and a hidden layer connected to the input layer, wherein: the input layer includes a plurality of input nodes, each input node corresponds to a hidden node connected to the hidden layer, and the second input module includes: A sorting module, used for arranging the device panel data in chronological order to obtain a device panel data sequence; A sub-input module, used for inputting the device panel data sequence into each input node in sequence; The processing module is used to process the output data of the previous hidden node and the device panel data of the corresponding input node for each hidden node to obtain output data; specifically, the processing module obtains the output data h(t) of the current hidden node through the following formula: ； Where: h(t-1) is the output data of the previous hidden node, X(t) is the device panel data of the input node corresponding to the current hidden node; f is the activation function, W is the first shared parameter, U is the second shared parameter, and b is the third shared parameter; An output module, used for taking the output data of the last hidden node as the second feature data; The number of input nodes and hidden nodes in the generalized model is variable, and the output module uses the output data of the last device panel data in the device panel data sequence as the second feature data; The memory model is a machine learning model.

5. The device according to claim 4, characterized in that The output layer is fully connected using the following formula: ； Wherein: y is the probability of device access loss, X1 is the first feature data, X2 is the second feature data, W1 is the parameter corresponding to the first feature data, W2 is the parameter corresponding to the second feature data; σ is the sigmoid activation function, and c is the bias term.

6. The device according to claim 4, characterized in that The device also includes: A pre-acquisition module is used to respectively acquire the equipment cross-section data of the historical equipment to be identified when the product is used, the historical equipment panel data refreshed over time, and the historical equipment label data; A pre-input module, used to initialize various parameters, input the device cross-section data of historical devices into the memory model of the device access loss model, and input the device panel data of historical devices into the generalized model of the device access loss model; A calculation module, used to calculate a loss function based on the output result of the device access loss model and historical device label data; The updating module is used to perform reverse operation based on the loss function to update parameters.

7. An electronic device comprising: processor; as well as A memory storing computer executable instructions which, when executed, cause the processor to perform the method according to any one of claims 1-3.

8. A computer-readable storage medium, wherein: The computer-readable storage medium stores one or more programs, and when the one or more programs are executed by a processor, the method according to any one of claims 1 to 3 is implemented.

Images