Method and system for analyzing and detecting advice work order pressure reduction rate influence factors

Abstract

The invention relates to the field of neural network analysis, in particular to an opinion work order pressure reduction rate influence factor analysis and detection method and system, and the method comprises the steps: obtaining influence factor parameters related to the pressure reduction rate of an opinion work order in each time period, constructing a parameter measurement value sequence, and a time sequence analysis model is combined with a neural network model to determine the weight priority of each influencing element parameter, and the element parameter which has the maximum influence on the commanded work order pressure reduction rate is discriminated. According to the method, the statistical characteristics of the time sequence analysis model and the learning ability of the neural network are combined, the abnormal value can be identified and processed more effectively, the sequence imported into the model is more general, and therefore it is guaranteed that the result exported by the model is more accurate.

Description

Technical Field

[0001] The present invention relates to the field of neural network analysis, and more specifically to a method and system for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders. Background Art

[0002] The number of opinion work orders is an important indicator to measure the service quality and customer satisfaction of power supply companies. It refers to the work documents generated after customers report problems or make opinions and suggestions by calling the power supply service hotline of various channels. The high number of work orders reflects that there are many problems with the service in the area. Processing a large number of work orders requires more manpower, material resources and time, which increases the pressure on employees, irrational resource allocation, increases operating costs, and affects the economic benefits of the enterprise; too many work orders externally may cause congestion in the processing process, resulting in reduced overall service efficiency, and failure to solve customer problems in a timely and effective manner, forming a vicious circle, resulting in poor customer experience, reduced satisfaction, and affecting the image and reputation of the company. At this stage, it is necessary to quickly locate the parameters of the factors that have the greatest impact on the number of opinion tickets. The existing technology mainly uses data survey table analysis and grading to gradually confirm the weighted influencing factors with the largest final proportion. However, there are disadvantages: 1) missing, incorrect or inconsistent data may lead to inaccurate analysis results and affect the reliability of the conclusions; 2) the survey process may be affected by the subjective opinions of the investigators or respondents, resulting in non-objective data; 3) if the sample is not selected properly, it may not represent the population, resulting in poor extrapolation of the conclusions; 4) table analysis is often static and cannot reflect the dynamic changes and trends of time series data.

[0003] Therefore, the present invention proposes a method and system for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders. Summary of the invention

[0004] In order to solve the shortcomings existing in the prior art, the present invention provides a method for analyzing and detecting factors affecting the reduction rate of opinion work orders, obtains the influencing factor parameters associated with the reduction rate of opinion work orders in each time period, constructs a parameter measurement value sequence, and uses a time series analysis model combined with a neural network model to determine the weight priority of each influencing factor parameter, and identifies the factor parameter that has the greatest impact on the reduction rate of opinion work orders.

[0005] The present invention adopts the following technical solution.

[0006] The present invention proposes a method for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders, comprising the following steps:

[0007] S1: Data sampling is used to obtain the influencing factor parameters associated with the pressure reduction rate of opinion work orders in each period, and a parameter measurement value sequence is constructed;

[0008] S2: Use a polynomial fitting algorithm to establish the relationship between the numerical points in the parameter measurement value sequence, and select the numerical points in the parameter measurement value sequence through the fitting function result to generate a new parameter measurement value sequence;

[0009] S3: Combine the time series analysis model to calculate the trend value of the parameter measurement sequence obtained in S2;

[0010] S4: Use the N trend values ​​obtained to import into the deep neural network learning model to calculate and obtain the weight priority of the influencing parameters and export the results.

[0011] In a further preferred embodiment, the polynomial fitting algorithm in S2 establishes the relationship between the numerical points in the parameter measurement sequence. The specific workflow is as follows:

[0012] Assume that the total number of influencing factor parameters is d, arbitrarily select the tth influencing factor parameter C from the influencing factor parameters for analysis, and obtain h measured values ​​of the influencing factor parameter C based on h different time points, that is, assume that there are h time periods T1, T2, …, T h and the corresponding measured values ​​C1, C2, ..., C h , construct the parameter measurement value sequence as C = {C1, C2, ..., C j ,…,C h}, according to the polynomial fitting algorithm, the j-th time period T j The measured value C j The high-order polynomial expression of is:

[0013]

[0014] Where n represents the highest order of the high-order polynomial expression, a0, a1, a2, ..., a n The parameters to be estimated are expressed as high-order polynomial expressions. The high-order polynomial expressions corresponding to the parameter measurement value sequence constructed by h measurement values ​​obtained in h different time periods are expressed by a matrix E, denoted as E=ZY.

[0015] In a further preferred embodiment, the matrix E is an h×1 column vector representing parameter measurement values;

[0016]

[0017] The matrix Z is an h×(n+1) time period matrix, where n represents the highest order of the high-order polynomial expression;

[0018]

[0019] The matrix Y is a (n+1)×1 coefficient column vector, indicating the parameters to be estimated of the high-order polynomial expression;

[0020]

[0021] Based on the matrix, the least squares algorithm is used to obtain the equation for the parameters to be estimated for the polynomial, which is:

[0022] E=(Z T Z) -1 Z T E;

[0023] The estimated parameters a0, a1, a2, ..., a of the polynomial obtained by solving the equation n Substitute back into the original expression In the above equation, we get the specific fitting polynomial equation model.

[0024] In a further preferred embodiment, the workflow of selecting numerical points in the parameter measurement value sequence by fitting function results in S2 is as follows:

[0025] Use the fitted polynomial equation to calculate the parameter forecast values ​​of all time points. For the fitted polynomial equation:

[0026] Among them, a0, a1, a2, ..., a n are the estimated parameters of the polynomial obtained by solving the above equation,

[0027] Prediction value pred (T j ) is calculated based on the fitted polynomial equation at time point T j The predicted value of the influencing factor parameter C;

[0028] Set a tolerance range to determine whether the actual measured value is consistent with the predicted value calculated under the fitted model.

[0029] In a further preferred embodiment, the tolerance is taken as a multiple of the standard deviation.

[0030] Further preferred solution, for each time period point T j , check the measured value of parameter C j and parameter prediction value l pred (T j ) is:

[0031] ER=|C j -l pred (T j )|;

[0032] If ER is greater than the set tolerance, delete the parameter measurement value of the time period; if ER is less than or equal to the set tolerance, retain the parameter measurement value of the time period and generate a new parameter measurement value sequence:

[0033] H represents the number of measurements in the new parameter measurement sequence, Represents the sth parameter measurement value in the new parameter measurement value sequence.

[0034] In a further preferred embodiment, S3 specifically includes:

[0035] The new parameter measurement value sequence constructed is: The s-1th parameter measurement value and the sth parameter measurement value The time period between s -T s-1 , T s Expressed as and The total time period, T s-1 Expressed as and The total time period between, then after the next interval, the sth parameter measurement value The single cumulative value of for:

[0036]

[0037] in, yes A single accumulation of yes Single accumulation of

[0038] The time series analysis model is:

[0039]

[0040] in, express , β and γ are the discrimination coefficients.

[0041] Then after N intervals, the sth parameter measurement value After N times, the cumulative value is The trend value of

[0042] In a further preferred embodiment, the step of importing the obtained N trend values ​​into the deep neural network learning model in S4 specifically includes:

[0043] Generate the corresponding trend values ​​of the H parameter measurement values ​​in the parameter measurement value sequence to construct the trend value sequence, and the trend value sequence is And the trend value sequence is introduced into the deep neural network learning model to obtain the generated derived sequence A'(N);

[0044] The generated derived sequence A'(N) is re-introduced into the deep neural network learning model to obtain the secondary generated sequence A"(N).

[0045] In a further preferred embodiment, the operation in S4 obtains the weight priority of the influencing parameter and derives the result including:

[0046] Arbitrarily select the weight priority W of the tth influencing factor parameter C among the influencing factor parameters i for:

[0047]

[0048] in, represents the weight coefficient of the g-th convolutional network layer in the deep neural network learning model, p tg represents the activation function of the g-th convolutional network layer, and the number of convolutional network layers is G;

[0049] p tg The activation function of the g-th convolutional network layer is:

[0050]

[0051] Among them, τ g represents the expansion rate of the g-th convolutional network layer, B g Represents the convolution kernel parameters of the g-th convolutional network layer.

[0052] The present invention also provides a system for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders, and uses the method for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders, including:

[0053] Data integration platform, factor parameter preprocessing module, factor parameter weight analysis module;

[0054] The data integration platform is used to sample data to obtain the influencing factor parameters associated with the pressure reduction rate of opinion work orders in each time period, and to construct a parameter measurement value sequence;

[0055] The element parameter preprocessing module is used to establish the relationship between the numerical points in the parameter measurement value sequence by using a polynomial fitting algorithm, and select the numerical points in the parameter measurement value sequence by using the fitting function result to generate a new parameter measurement value sequence;

[0056] The factor parameter weight analysis module is used to import the trend value into the deep neural network learning model to calculate and obtain the weight priority of the influencing parameters and export the results.

[0057] The beneficial effect of the present invention is that, compared with the prior art, the present invention provides a method for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders, obtains the influencing factor parameters associated with the pressure reduction rate of opinion work orders in each time period, constructs a parameter measurement value sequence, and uses a time series analysis model combined with a neural network model to determine the weight priority of each influencing factor parameter, and identifies the factor parameter that has the greatest impact on the pressure reduction rate of opinion work orders. The present invention can predict future trends and patterns based on historical influencing factor parameters and improve the accuracy of decision-making. The combination of the time series analysis model and the neural network model can enhance the calculation accuracy of the factor parameter weights. Combining the statistical characteristics of the time series analysis model with the learning ability of the neural network can more effectively identify and process outliers, making the sequence imported into the model more general, thereby ensuring that the results derived from the model are more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a flow chart of the method for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders in the present invention. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the scope of protection of the present invention.

[0060] refer to Figure 1 The embodiment of the present invention provides a method for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders, comprising the following steps:

[0061] S1: Data sampling is used to obtain the influencing factor parameters associated with the pressure reduction rate of opinion work orders in each time period, and a parameter measurement value sequence is constructed.

[0062] In the embodiment of the present invention, there are three types of influencing factor parameters associated with the voltage reduction rate of opinion work orders: the first type of opinion work orders classified by acceptance channels include 95598 channels, grid channels, 12398 energy supervision bureau hotline channels, and 12345 municipal hotline channels; the second type of opinion work orders classified by business type include electricity prices and charges, power supply quality, wrong text messages, service behaviors, business expansion applications, and grid stability; the third type of opinion work orders classified by power supply quality include load changes, operating conditions of power supply equipment, voltage stability, equipment failures, and power market dynamics. Based on the influencing parameters obtained by understanding, several influencing factor parameters are selected, including power supply quality, load changes, operating conditions of power supply equipment, voltage stability, equipment failures, and power market dynamics.

[0063] S2: Use a polynomial fitting algorithm to establish the relationship between the numerical points in the parameter measurement value sequence, and select the numerical points in the parameter measurement value sequence through the fitting function results to generate a new parameter measurement value sequence.

[0064] Further, the polynomial fitting algorithm establishes the relationship between the numerical points in the parameter measurement value sequence. The specific workflow is as follows:

[0065] Assume that the total number of influencing factor parameters is d, arbitrarily select the tth influencing factor parameter C from the influencing factor parameters for analysis, and obtain h measured values ​​of the influencing factor parameter C based on h different time points, that is, assume that there are h time periods T1, T2, …, T h and the corresponding measured values ​​C1, C2, ..., C h , construct the parameter measurement value sequence as C = {C1, C2, ..., C j ,…,C h}, according to the polynomial fitting algorithm, the j-th time period T j The measured value C j The high-order polynomial expression of is:

[0066]

[0067] Where n represents the highest order of the high-order polynomial expression, a0, a1, a2, ..., a n The parameters to be estimated are expressed as high-order polynomial expressions. The high-order polynomial expressions corresponding to the parameter measurement value sequence constructed by h measurement values ​​obtained in h different time periods are expressed by a matrix E, denoted as E=ZY.

[0068] Specifically, the matrix E is an h × 1 column vector, representing the parameter measurement value;

[0069]

[0070] The matrix Z is an h×(n+1) time period matrix, where n represents the highest order of the high-order polynomial expression;

[0071]

[0072] The matrix Y is a (n+1)×1 coefficient column vector, indicating the parameters to be estimated for the high-order polynomial expression;

[0073]

[0074] Based on the matrix, the least squares algorithm is used to obtain the equation for the parameters to be estimated for the polynomial, which is:

[0075] E=(Z T Z) -1 ZT E.

[0076] The estimated parameters a0, a1, a2, ..., a of the polynomial obtained by solving the equation n Substitute back into the original expression In the above equation, we get the specific fitting polynomial equation model.

[0077] The obtained specific fitting polynomial equation model is used to select numerical points in the parameter measurement value sequence through the fitting function result, specifically including:

[0078] Use the fitted polynomial equation to calculate the parameter forecast values ​​of all time points. For the fitted polynomial equation:

[0079] Among them, a0, a1, a2, ..., a n are the estimated parameters of the polynomial obtained by solving the above equation,

[0080] Prediction value L pred (T j ) is calculated based on the fitted polynomial equation at time point T j The predicted value of the influencing factor parameter C.

[0081] Set a tolerance range to determine whether the actual measured value meets the predicted value calculated under the fitted model. Usually the tolerance can be set according to the specific situation, such as a multiple of the standard deviation.

[0082] For each time period T j , check the measured value of parameter C j and parameter prediction value l pred (T j ) is:

[0083] ER=|C j -l pred (T j )|;

[0084] If ER is greater than the set tolerance, it is judged that the discrete amplitude of the parameter measurement value is large, and the parameter measurement value of this time period is deleted; if ER is less than or equal to the set tolerance, it is judged that the discrete amplitude of the parameter measurement value is small, and the parameter measurement value of this time period is retained, and a new parameter measurement value sequence is generated as follows: H represents the number of measurements in the new parameter measurement sequence, Represents the sth parameter measurement value in the new parameter measurement value sequence.

[0085] S3: Combine the time series analysis model to calculate the trend value of the parameter measurement value sequence obtained in step 2.

[0086] The new parameter measurement value sequence constructed is: The s-1th parameter measurement value and the sth parameter measurement value The time period between s -T s-1 , T s Expressed as and The total time period, T s-1 Expressed as and The total time period between, then after the next interval, the sth parameter measurement value The single cumulative value of for:

[0087]

[0088] in, yes A single accumulation of yes Single accumulation of

[0089] The time series analysis model is:

[0090]

[0091] in, express , β and γ are the discrimination coefficients.

[0092] Then after N intervals, the sth parameter measurement value After N times, the cumulative value is The trend value of

[0093] S4: Use the N trend values ​​obtained to import into the deep neural network learning model to calculate and obtain the weight priority of the influencing parameters and export the results.

[0094] Generate the corresponding trend values ​​of the H parameter measurement values ​​in the parameter measurement value sequence to construct the trend value sequence, and the trend value sequence is And the trend value sequence is introduced into the deep neural network learning model to obtain the generated derived sequence A'(N);

[0095] The generated derived sequence A'(N) is re-introduced into the deep neural network learning model to obtain the secondary generated sequence A"(N).

[0096] Arbitrarily select the weight priority W of the tth influencing factor parameter C among the influencing factor parameters i for:

[0097]

[0098] in, represents the weight coefficient of the g-th convolutional network layer in the deep neural network learning model, p tg Represents the activation function of the g-th convolutional network layer, and the number of convolutional network layers is G.

[0099] p tg The activation function of the g-th convolutional network layer is:

[0100]

[0101] Among them, τ g represents the expansion rate of the g-th convolutional network layer, B g Represents the convolution kernel parameters of the g-th convolutional network layer.

[0102] The above processing flow is adopted for d influencing factor parameters in turn, and the weight priority values ​​of the d influencing factor parameters finally obtained constitute the weight priority value W. The d influencing factor parameters are arranged in sequence from small to large according to the weights of the influencing factor parameters. The influencing factor parameter with the highest weight priority is at the end of the sequence. At the same time, the influencing factor parameter with the highest weight priority is also the factor that has the greatest impact on the pressure reduction rate of opinion work orders.

[0103] The embodiment of the present invention also provides a system for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders, using the method for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders, including a data integration platform, a factor parameter preprocessing module, and a factor parameter weight analysis module;

[0104] The data integration platform is used to sample data to obtain the influencing factor parameters associated with the pressure reduction rate of opinion work orders in each time period, and to construct a parameter measurement value sequence;

[0105] The element parameter preprocessing module is used to establish the relationship between the numerical points in the parameter measurement value sequence by using a polynomial fitting algorithm, and select the numerical points in the parameter measurement value sequence by using the fitting function result to generate a new parameter measurement value sequence;

[0106] The factor parameter weight analysis module is used to import the trend value into the deep neural network learning model to calculate and obtain the weight priority of the influencing parameters and export the results.

[0107] The beneficial effect of the present invention is that, compared with the prior art, the present invention provides a method for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders, obtains the influencing factor parameters associated with the pressure reduction rate of opinion work orders in each time period, constructs a parameter measurement value sequence, and uses a time series analysis model combined with a neural network model to determine the weight priority of each influencing factor parameter, and identifies the factor parameter that has the greatest impact on the pressure reduction rate of opinion work orders. The present invention can predict future trends and patterns based on historical influencing factor parameters and improve the accuracy of decision-making. The combination of the time series analysis model and the neural network model can enhance the calculation accuracy of the factor parameter weights. Combining the statistical characteristics of the time series analysis model with the learning ability of the neural network can more effectively identify and process outliers, making the sequence imported into the model more general, thereby ensuring that the results derived from the model are more accurate.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders, characterized in that: The following steps are involved: S1: Data sampling is used to obtain the influencing factor parameters associated with the pressure reduction rate of opinion work orders in each period, and a parameter measurement value sequence is constructed; S2: Use a polynomial fitting algorithm to establish the relationship between the numerical points in the parameter measurement value sequence, and select the numerical points in the parameter measurement value sequence through the fitting function result to generate a new parameter measurement value sequence; S3: Combine the time series analysis model to calculate the trend value of the parameter measurement sequence obtained in S2; S4: Use the N trend values ​​obtained to import into the deep neural network learning model to calculate and obtain the weight priority of the influencing parameters and export the results.

2. The method for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders according to claim 1 is characterized in that: The specific workflow of the polynomial fitting algorithm in S2 to establish the relationship between the numerical points in the parameter measurement sequence is as follows: Assume that the total number of influencing factor parameters is d, arbitrarily select the tth influencing factor parameter C from the influencing factor parameters for analysis, and obtain h measured values ​​of the influencing factor parameter C based on h different time points, that is, assume that there are h time periods T1, T2, …, T h and the corresponding measured values ​​C1, C2, ..., C h , construct the parameter measurement value sequence as C = {C1, C2, ..., C j ,…,C h }, according to the polynomial fitting algorithm, the j-th time period T j The measured value C j The high-order polynomial expression of is: Where n represents the highest order of the high-order polynomial expression, a0, a1, a2, ..., a n The parameters to be estimated are expressed as high-order polynomial expressions. The high-order polynomial expressions corresponding to the parameter measurement value sequence constructed by h measurement values ​​obtained in h different time periods are expressed by a matrix E, denoted as E=ZY.

3. The method for analyzing and detecting factors affecting pressure reduction rate of opinion work orders according to claim 2, characterized in that: The matrix E is an h×1 column vector, representing parameter measurement values; The matrix Z is an h×(n+1) time period matrix, where n represents the highest order of the high-order polynomial expression; The matrix Y is a (n+1)×1 coefficient column vector, indicating the parameters to be estimated of the high-order polynomial expression; Based on the matrix, the least squares algorithm is used to obtain the equation for the parameters to be estimated for the polynomial, which is: E=(Z T (W) -1 With T E; The estimated parameters a0, a1, a2, ..., a of the polynomial obtained by solving the equation n Substitute back into the original expression In the above equation, we get the specific fitting polynomial equation model.

4. The method for analyzing and detecting factors affecting pressure reduction rate of opinion work orders according to claim 1, characterized in that: The workflow for selecting numerical points in the parameter measurement value sequence by fitting function results described in S2 is as follows: Use the fitted polynomial equation to calculate the parameter forecast values ​​of all time points. For the fitted polynomial equation: Among them, a0, a1, a2, ..., a n are the estimated parameters of the polynomial obtained by solving the above equation, Prediction value pred (T j ) is calculated based on the fitted polynomial equation at time point T j The predicted value of the influencing factor parameter C; Set a tolerance range to determine whether the actual measured value is consistent with the predicted value calculated under the fitted model.

5. The method for analyzing and detecting factors affecting pressure reduction rate of opinion work orders according to claim 4 is characterized in that: The tolerance is taken as a multiple of the standard deviation.

6. A method for analyzing and detecting factors affecting pressure reduction rate of opinion work orders according to claim 5, characterized in that: For each time period T j , check the measured value of parameter C j and parameter prediction value l pred (T j ) is: IS=|C j -l pred (T j )|; If ER is greater than the set tolerance, delete the parameter measurement value of the time period; if ER is less than or equal to the set tolerance, retain the parameter measurement value of the time period and generate a new parameter measurement value sequence: H represents the number of measurements in the new parameter measurement sequence, Represents the sth parameter measurement value in the new parameter measurement value sequence.

7. The method for analyzing and detecting factors affecting pressure reduction rate of opinion work orders according to claim 1, characterized in that: The S3 specifically includes: The new parameter measurement value sequence constructed is: The s-1th parameter measurement value and the sth parameter measurement value The time period between s -T s-1 , T s Expressed as and The total time period, T s-1 Expressed as and The total time period between, then after the next interval, the sth parameter measurement value The single cumulative value of for: in, yes A single accumulation of yes Single accumulation of The time series analysis model is: in, express , β and γ are the discrimination coefficients. Then after N intervals, the sth parameter measurement value After N times, the cumulative value is The trend value of 8. The method for analyzing and detecting factors affecting pressure reduction rate of opinion work orders according to claim 1, characterized in that: The step of importing the obtained N trend values ​​into the deep neural network learning model in S4 specifically includes: Generate the corresponding trend values ​​of the H parameter measurement values ​​in the parameter measurement value sequence to construct the trend value sequence, and the trend value sequence is And the trend value sequence is introduced into the deep neural network learning model to obtain the generated derived sequence A'(N); The generated derived sequence A'(N) is re-introduced into the deep neural network learning model to obtain the secondary generated sequence A"(N).

9. The method for analyzing and detecting factors affecting pressure reduction rate of opinion work orders according to claim 1, characterized in that: The operation in S4 obtains the weight priority of the influencing parameter and derives the result including: Arbitrarily select the weight priority W of the tth influencing factor parameter C among the influencing factor parameters i for: in, represents the weight coefficient of the g-th convolutional network layer in the deep neural network learning model, p tg represents the activation function of the g-th convolutional network layer, and the number of convolutional network layers is G; p tg The activation function of the g-th convolutional network layer is: Among them, τ g represents the expansion rate of the g-th convolutional network layer, B g Represents the convolution kernel parameters of the g-th convolutional network layer.

10. A system for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders, using a method for analyzing and detecting factors affecting the pressure reduction rate of opinion work orders as claimed in any one of claims 1 to 9, comprising: Data integration platform, factor parameter preprocessing module, factor parameter weight analysis module; The data integration platform is used to sample data to obtain the influencing factor parameters associated with the pressure reduction rate of opinion work orders in each time period, and to construct a parameter measurement value sequence; The element parameter preprocessing module is used to establish the relationship between the numerical points in the parameter measurement value sequence by using a polynomial fitting algorithm, and select the numerical points in the parameter measurement value sequence by using the fitting function result to generate a new parameter measurement value sequence; The factor parameter weight analysis module is used to import the trend value into the deep neural network learning model to calculate and obtain the weight priority of the influencing parameters and export the results.

Images