Liquid milk product quality problem complaint prediction method and device, medium and product

Abstract

The invention discloses a liquid milk product quality problem complaint prediction method and device, a medium and a product, and relates to the field of complaint prediction. The method comprises the following steps: acquiring quality control point list data; inputting the quality control point list data into the liquid milk product complaint prediction model, and outputting predicted complaint data; the liquid milk product complaint prediction model is obtained by training a constructed full-connection neural network based on a historical complaint causal relationship data set of each process link; the full-connection neural network comprises an input layer, a hidden layer and an output layer which are connected in sequence; the historical complaint causal relationship data set is coding sequence data determined by performing information matching on historical complaint information, historical inspection problem information and historical liquid milk product quality control point list data according to a control point parameter setting mode; according to the invention, quality problem complaint prediction can be realized, and potential problem products can be intervened in advance.

Description

Technical Field

[0001] The present application relates to the field of complaint prediction, and in particular to a method, device, medium and product for predicting complaints about quality issues of liquid milk products. Background Art

[0002] Liquid milk production is a multi-link, uninterrupted processing process, and a large number of quality-influencing factors in each link have an impact on the quality of the final product. At present, most dairy companies have sorted out the factors that may affect the quality of liquid milk products in their respective production lines, namely quality control points. However, the inspection information of these quality control points remains in the table, and the better practice is to apply simple statistical results to predict the effect of inspection information on product quality. As a complex production system, the quality problems that often lead to customer complaints are not caused by the abnormal state of a quality control point, and not all combinations of abnormal quality control points can have precedents to follow. The current quality control method cannot properly handle and utilize the abnormal quality control point information found in multiple quality inspections over a period of time; at the same time, the customer complaint information received has not been organically combined with the problems found during the inspection process, and the function of preventing future quality problem complaints by analyzing historical records has not been realized. Summary of the invention

[0003] The purpose of this application is to provide a method, device, medium and product for predicting complaints about quality problems of liquid milk products, which can realize the prediction of complaints about quality problems and intervene in potential problem products in advance.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a method for predicting complaints about liquid milk product quality issues, comprising:

[0006] Acquire quality control point list data; the quality control point list data includes: corresponding control parameters of each sub-process in the process link during the liquid milk production process; the process link includes: raw milk pre-treatment, milk filling and milk packaging; the control parameters include: buffer chain lubrication level, main compressed air pressure and straw heating temperature;

[0007] The quality control point list data is input into the liquid milk product complaint prediction model, and the predicted complaint data is output; the predicted complaint data includes: complaint type and complaint probability; the liquid milk product complaint prediction model is obtained by training a constructed fully connected neural network based on the historical complaint causal relationship data set of each process link; the liquid milk product complaint prediction model includes: a raw milk pre-processing complaint prediction sub-model, a milk filling complaint prediction sub-model and a milk packaging complaint prediction sub-model; the fully connected neural network includes an input layer, a hidden layer and an output layer connected in sequence; the historical complaint causal relationship data set is a coded sequence data determined by matching historical complaint information, historical inspection problem information and historical liquid milk product quality control point list data according to the set control point parameter setting method.

[0008] Optionally, the method for determining the liquid milk product complaint prediction model specifically includes:

[0009] Obtain historical complaint causal relationship dataset;

[0010] Dividing the historical complaint causal relationship dataset into a training set and a test set according to a set ratio;

[0011] Build a fully connected neural network;

[0012] The training set is input into the hidden layer through the input layer, and the weight parameters of the hidden layer are trained and updated by using the stochastic gradient descent method and the momentum algorithm with the objective of minimizing the objective function, so as to obtain the trained hidden layer; the objective function is determined according to the deviation between the output value of the hidden layer and the actual value of the historical complaint information;

[0013] The test set is used to adjust the parameters of the trained hidden layer to obtain an adjusted hidden layer; the parameters include: learning rate and momentum parameter;

[0014] The output layer is used to use the tanh function to perform nonlinear relationship fitting, and uses the loss function and the softmax function to determine the historical complaint data according to the output value of the adjusted hidden layer;

[0015] Determining a trained fully connected neural network according to the input layer, the adjusted hidden layer and the output layer;

[0016] The trained fully connected neural network is determined as the liquid milk product complaint prediction model.

[0017] Optionally, the obtaining of a historical complaint causal relationship dataset specifically includes:

[0018] Obtain historical inspection problem information, historical complaint information, and historical liquid milk product quality control point list data;

[0019] According to the historical complaint information and the historical inspection problem information, information matching is performed to obtain first matching relationship information;

[0020] Perform information matching according to the first matching relationship information, the historical inspection problem information, and the historical liquid milk product quality control point list data to obtain second matching relationship information;

[0021] According to the historical complaint information and the second matching relationship information, the coding sequence data is determined in accordance with the setting control point parameter setting method;

[0022] The historical complaint causal relationship dataset is determined based on the coding sequence data.

[0023] Optionally, the liquid milk product quality complaint prediction method further includes:

[0024] Comparing the complaint prediction data with a preset threshold value to obtain a comparison result; the preset threshold value includes: a first threshold value and a second threshold value; the first threshold value is less than the second threshold value;

[0025] If the predicted data is greater than the first threshold but less than the second threshold, a prompt message is issued;

[0026] If the predicted data is greater than or equal to the second threshold, a warning message is issued; the warning message is used to prompt the staff to conduct product testing.

[0027] Optionally, the loss function adopts a cross entropy loss function;

[0028] The expression of the loss function is:

[0029]

[0030] Where L is the loss function; N is the total number of data in the historical complaint causal relationship dataset; i is the sequence number; j is the label of the complaint type in the complaint prediction data; c ij is a constant value indicating whether the output value of the i-th group of data is really a complaint of the j-th type; p ij For complaint prediction data, the probability of complaint;

[0031] p ij It is obtained by processing with the softmax function, and the calculation formula is:

[0032]

[0033] y ij is the output value of the adjusted hidden layer.

[0034] Optionally, the liquid milk product quality complaint prediction method further includes:

[0035] A knowledge graph method is used to perform visualization based on the predicted complaint data.

[0036] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned method for predicting complaints about quality issues of liquid milk products.

[0037] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for predicting complaints about quality issues of liquid milk products.

[0038] In a fourth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned method for predicting complaints about quality issues of liquid milk products.

[0039] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0040] The present application provides a method, device, medium and product for predicting complaints about quality problems of liquid milk products, by obtaining quality control point list data; inputting the quality control point list data into a liquid milk product complaint prediction model, and outputting predicted complaint data; the predicted complaint data includes: complaint type and complaint probability; the liquid milk product complaint prediction model is obtained by training a constructed fully connected neural network based on a historical complaint causal relationship data set of each process link; the historical complaint causal relationship data set is obtained by matching historical complaint information, historical inspection problem information and historical liquid milk product quality control point list data, and determining coded sequence data according to a set control point parameter setting method; the present application takes the realization of "prediction of complaints about quality problems of liquid milk products" as the core goal, provides a comprehensive solution, improves the comprehensive utilization capability of quality information of liquid milk production enterprises, reduces the complexity of workers' work, realizes the quality complaint prediction function, and provides technical support for enterprises to intervene in potential problem products in advance and prevent them from entering the market; thereby, it is possible to realize the prediction of quality problem complaints and intervene in potential problem products in advance. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 This is a flow chart of a method for predicting complaints about liquid milk product quality issues in one embodiment of the present application;

[0043] Figure 2 A diagram of the overall method framework for constructing a historical complaint causal relationship dataset provided in an embodiment of the present application;

[0044] Figure 3 A schematic diagram of a liquid milk product complaint prediction model provided in one embodiment of the present application;

[0045] Figure 4 A schematic diagram of the interaction logic between the display terminal and the database in the task of predicting complaints about liquid milk quality problems provided in an embodiment of the present application;

[0046] Figure 5 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0048] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0049] like Figure 1 As shown, the present application provides a method for predicting complaints about liquid milk product quality issues, the method comprising:

[0050] Get the quality control point list data. The quality control point list data includes: the corresponding control parameters of each sub-process in the process link during the liquid milk production process; the process link includes: raw milk pre-treatment, milk filling and milk packaging; the control parameters include: buffer chain lubrication level, main compressed air pressure and straw heating temperature.

[0051] The quality control point list data is input into the liquid milk product complaint prediction model, and the predicted complaint data is output. The predicted complaint data includes: complaint type and complaint probability; the liquid milk product complaint prediction model is based on the historical complaint causal relationship data set of each process link, and the constructed fully connected neural network is trained; the liquid milk product complaint prediction model includes: raw milk pre-processing complaint prediction sub-model, milk filling complaint prediction sub-model and milk packaging complaint prediction sub-model; the fully connected neural network includes an input layer, a hidden layer and an output layer connected in sequence; the historical complaint causal relationship data set is determined by matching historical complaint information, historical inspection problem information and historical liquid milk product quality control point list data, and setting the control point parameters according to the set control point parameter setting method. The coded sequence data.

[0052] In an exemplary embodiment, a method for determining a liquid milk product complaint prediction model specifically includes:

[0053] Obtain a historical complaint causal relationship dataset; divide the historical complaint causal relationship dataset into a training set and a test set according to a set ratio.

[0054] Construct a fully connected neural network; input the training set into the hidden layer through the input layer, and use the stochastic gradient descent method and momentum algorithm to train and update the weight parameters of the hidden layer with the goal of minimizing the objective function to obtain the trained hidden layer; the objective function is determined based on the deviation between the output value of the hidden layer and the actual value of the historical complaint information.

[0055] The test set is used to adjust the parameters of the trained hidden layer to obtain the adjusted hidden layer; the parameters include: learning rate and momentum parameter.

[0056] The output layer is used to fit the nonlinear relationship using the tanh function, and uses the loss function and the softmax function to determine the historical complaint data based on the output value of the adjusted hidden layer.

[0057] According to the input layer, the adjusted hidden layer and the output layer, a trained fully connected neural network is determined; and the trained fully connected neural network is determined as a liquid milk product complaint prediction model.

[0058] Specifically, the loss function used is the cross entropy loss function.

[0059] The expression of the loss function is:

[0060]

[0061] Where L is the loss function; N is the total number of data in the historical complaint causal relationship dataset; i is the sequence number; j is the label of the complaint type in the complaint prediction data; c ijis a constant value indicating whether the output value of the i-th group of data is really a complaint of the j-th type; p ij Predict the probability of complaint in the complaint data.

[0062] p ij It is obtained by processing with the softmax function, and the calculation formula is:

[0063]

[0064] y ij is the output value of the adjusted hidden layer.

[0065] In an exemplary embodiment, obtaining a historical complaint causal relationship dataset specifically includes:

[0066] Obtain historical inspection problem information, historical complaint information, and historical liquid milk product quality control point list data.

[0067] According to the historical complaint information and the historical inspection problem information, information matching is performed to obtain the first matching relationship information.

[0068] Information matching is performed based on the first matching relationship information, the historical inspection problem information, and the quality control point list data of historical liquid milk products to obtain the second matching relationship information.

[0069] According to the historical complaint information and the second matching relationship information, the coding sequence data is determined according to the set control point parameter setting method.

[0070] A historical complaint causality dataset was determined from coded sequence data.

[0071] In an exemplary embodiment, the method for predicting complaints about liquid milk product quality issues further includes:

[0072] The complaint prediction data is compared with a preset threshold to obtain a comparison result; the preset threshold includes: a first threshold and a second threshold; the first threshold is smaller than the second threshold.

[0073] If the predicted data is greater than the first threshold but less than the second threshold, a prompt message is issued.

[0074] If the predicted data is greater than or equal to the second threshold, a warning message is issued; the warning message is used to prompt the staff to conduct product testing.

[0075] In an exemplary embodiment, the method for predicting complaints about liquid milk product quality issues further includes:

[0076] The knowledge graph method is used to perform visualization based on the predicted complaint data.

[0077] The present application also provides an application scenario, which applies the above-mentioned liquid milk product quality problem complaint prediction method. Specifically: The liquid milk product quality problem complaint prediction method provided in this embodiment can be applied in the quality problem complaint prediction scenario with liquid pure milk products as the research object. The present application proposes a quality problem complaint prediction method that integrates neural network data processing technology and knowledge graph data display technology. Starting from the production line inspection history record and the complaint information history record, the method pre-processes and integrates the data of two different information sources, and predicts from the production process quality problems to the client complaint problems through the neural network. This method overcomes the problem of information not being interoperable between different departments and different data sources of dairy enterprises, and the low data utilization rate, and predicts the problem complaints that may occur after liquid milk products enter the market from the source. On this basis, the knowledge graph technology is used to develop a quality management human-computer interaction interface suitable for liquid milk production enterprises, forming a set of liquid milk product quality problem complaint prediction system, which overcomes the problem of labor-intensive and material-intensive quality management, and provides enterprises with a digital and intelligent quality problem prediction tool.

[0078] The production process of liquid milk involves multiple process links. From the milk truck unloading the raw milk from the ranch to the final packaging of the product, it goes through three major processes: raw milk pre-treatment, milk canning, and milk packaging. Each process contains many sub-processes and production equipment systems. Each production equipment system can realize one or more functions, which can be further divided into one or more quality control points. Each control point can be described by a specific standard or parameter. The state of the description is a factor that affects the quality of liquid milk products. For example: the packaging process includes the buffer chain process, and the buffer chain process includes the dry lubrication system. The dry lubrication system has a chain lubrication function. One of the quality control points of the chain lubrication function is the lubrication level. The standard / parameter describing the lubrication level is "the liquid level cannot be lower than 1 / 4". At a certain moment, whether the liquid level is lower than 1 / 4 becomes a factor that affects the quality of liquid milk products.

[0079] Based on the above knowledge, the "Liquid Milk Production Process Quality Control Standard Information Table" was designed, and Table 1 is an excerpt of the table. At the same time, by collecting quality inspection information, the "Inspection Problem Record Table" was obtained; by collecting customer complaint information, the "Complaint Information Record Table" was obtained. The comprehensive use of the information in these tables will enable customer complaints to be predicted in advance, so as to be controlled in advance.

[0080] Table 1 Excerpt from the quality control standard information table for liquid milk production process

[0081]

[0082]

[0083] Specifically, the operation steps of this method in practical application can be as follows:

[0084] Step 1: Information fusion to construct a historical complaint causal relationship dataset.

[0085] The information involved in the "Liquid Milk Production Process Quality Control Standard Information Form", "Inspection Problem Record Form" and "Complaint Information Record Form" is relatively scattered and needs to be considered comprehensively and integrated for use. The overall method framework for constructing a historical complaint causal relationship dataset is as follows: Figure 2 shown.

[0086] (1) The information in the "Complaint Information Record Form" and the "Inspection Problem Record Form" is matched. That is, the information in the historical complaint information is matched with the information in the historical inspection problem information.

[0087] The "Complaint Information Record Form" records the date on which each complaint occurred, the production date of the corresponding product batch number, and can roughly locate the process where the problem occurred (i.e., pre-processing problem / filling problem / packaging problem). The "Inspection Problem Record Form" records the process steps on the liquid milk production line that do not meet the quality control point standards and the process to which the step belongs, and also records the inspection time. For each record in the "Complaint Information Record Form", the "production date" information and "problem process" information are matched with the "inspection time" information and "belonging process" information in the "Inspection Problem Record Form" to obtain a set of corresponding relationships, that is, one complaint record corresponds to multiple problem information found during the inspection process.

[0088] (2) Information matching between the Inspection Problem Record Form and the Liquid Milk Production Process Quality Control Standard Information Form, that is, information matching between historical inspection problem information and historical liquid milk product quality control point list data.

[0089] For each complaint record, multiple inspection records matching it in the Inspection Problem Record Table are obtained in step 1.(1). The problems recorded in these inspection records may have caused the product to have the problem being complained about, and each inspection record also records the name of the quality control point that does not meet the description and the correct standard / parameter. By matching the "quality control point name" of each inspection record in the Inspection Problem Record Table with the "quality control point" information in the Liquid Milk Production Process Quality Control Standard Information Table, a new set of corresponding relationships can be obtained, that is, one complaint record corresponds to multiple quality control point abnormal information.

[0090] (3) Generation of historical complaint causal relationship dataset.

[0091] For each complaint record, the corresponding multiple abnormal quality control points that may cause this complaint are obtained in step 1. (2). The parameter values ​​of these quality control points are set to 1, and the parameters of other quality control points in the process are set to 0. The types of complaints (sour / swollen / bitter / caking) that appear are encoded as

[1000] ,

[0100] ,

[0010] ,

[0001] , respectively. A complaint causal relationship data can be generated for each complaint record. The data format is [0 or 1]*number(quality control point)→[0 or 1]*4. The collection of complaint causal relationship data generated by each complaint record is the "historical complaint causal relationship data set".

[0092] Step 2: Build and train a liquid milk product complaint prediction neural network model, that is, build a liquid milk product complaint prediction model.

[0093] The problem of predicting complaints about liquid milk products can be abstracted as a classification problem in traditional deep learning algorithms, with each of the four complaint types of sour / swollen / bitter / lumps being one category. A neural network classifier can be constructed by taking the data list of quality control points as input and the data list of complaint types as output. Since each process is relatively independent, the pre-processing, filling, and packaging processes are modeled separately; since the quality control points of each process are less than 200, they can be considered as small vector inputs, so a fully connected neural network is used for modeling. The model diagram is shown below: Figure 3 shown.

[0094] The network activation function uses the tanh function To fit the nonlinear relationship between quality control points and complaint issues, the loss function uses the cross entropy loss function commonly used in classification problems. where c ij Refers to whether the output value of the i-th group of data is really a j-type complaint, 1 for true and 0 for false; p ij The output value of the i-th group of data after network training is the probability of the j-th type of complaint. The probability value p ij Using the softmax function to process it, The network output is a probability value between 0 and 1.

[0095] The historical complaint causal relationship dataset is randomly divided into a 70% training set and a 30% test set, and the parameters of the neural network are updated using the stochastic gradient descent + momentum algorithm. For the training set, the deviation between the output value obtained by the neural network operation and the true value is ΔY. For the weight parameter W in the hidden layer of the neural network, its gradient value at the current output position is calculated. Update the momentum value of the weight at this output position And the new weight value w=w-ηv after gradient descent. Repeat the forward operation and reverse weight update of the neural network, debug the learning rate η and momentum parameter β, until the optimal fit effect is obtained. For new problems found during the inspection process, these problem information can be organized into the format of input data according to a certain time interval, and the neural network can be used for prediction, and the probability of each complaint problem occurring can be output, thereby realizing the prediction function. When it is predicted that a certain batch of products may cause a certain complaint problem, the testing engineer can conduct targeted product testing to prevent the problem products from entering the market.

[0096] Step 3: Design the human-computer interaction behavior of the liquid milk product complaint prediction process.

[0097] The knowledge graph technology is used to visualize the liquid milk complaint prediction process, and data interaction is carried out with the help of graph database and relational database technology, which facilitates the understanding and application of the liquid milk product complaint prediction model by quality staff in the liquid milk workshop.

[0098] (1) Input of quality inspection records of liquid milk process.

[0099] In view of the numerous entries and difficult to query "Liquid Milk Production Process Quality Control Standard Information Table", the knowledge graph technology is used to convert the table into a "Liquid Milk Production Process Quality Control Point Retrieval Map". The implementation method is to abstract the five-level information of process-sub-process-system-function-quality control point in the table into five-level nodes, where the first-level node is the parent node, containing process information (including pre-treatment / filling / packaging), the fifth-level node is the child node, containing quality control point information (such as lubrication level, etc.), and the middle nodes represent sub-processes, systems, and functions respectively. When entering the inspection record, each interface only displays one of the first-level nodes. Click the process to which the quality control point with the problem belongs to enter the sub-process interface; and select the sub-process, system, and function where the problem is located in turn until entering the quality control point interface. Click the abnormal quality control point, and a dialog box pops up, where you can enter the description of the abnormal state of the quality control point or the abnormal parameter value. This set of operations is used to replace the inspection information entry work using Excel tables.

[0100] (2) Information interaction in the liquid milk quality complaint prediction process.

[0101] Taking a certain time interval as a time unit, the input information of abnormal quality control points within this period is collected to carry out complaint problem prediction. When a description of the abnormal state of a quality control point or an abnormal parameter value is entered, the description information is stored in the graph database and the relational database at the same time. Among them, the graph database is a real-time update database, and its function is to ensure that the knowledge graph displays the latest status of the quality control points in each process link; the relational database is a storage function database, and each time an abnormal quality control point information is entered on the basis of retaining the original data, a new data record is added. The staff can click the "General Map" option to query the spatial relationship of the distribution of all abnormal quality control points in the time unit based on the graph database technology; at the same time, they can click the "Prediction" option to generate a prediction network input set and inspection problem report based on the relational database, and automatically calculate the probability of occurrence of four types of complaint problems through the trained neural network and display them on the interactive interface. Figure 4 The interaction logic between the display end and the database in the task of predicting complaints about liquid milk quality issues is demonstrated.

[0102] Step 4: Microservice development of liquid milk product complaint prediction system.

[0103] Based on the model constructed in the previous steps, the development and application of the liquid milk product quality problem complaint prediction system is realized based on the microservice architecture with front-end and back-end separation. The front-end user interface program is developed using the Node.js framework to realize the interactive viewing of the graph structure of product quality control points and the structured upload of abnormal conditions, and the back-end interface is called based on the Axios request library. The back-end microservice system framework is built using Spring CloudAlibaba. The complaint prediction microservice and neural network microservice are constructed using Java and Python languages ​​respectively. The former integrates the abnormal conditions of the quality control points input by the user, confirms its validity by interacting with the graph database Neo4j, and transmits the graph structure and abnormal information as a whole to the latter through the Feign component to realize the model operation. The operation results are returned to the complaint prediction microservice and delivered to the front-end program for display in the user interface.

[0104] The application process of this system can be triggered by manual clicks or automatically by the MES system. After the quality control point abnormality is found on the production line, the quality engineer will identify and locate the quality control point information step by step, and fill in the quality control point abnormality in the system according to the facts. After submitting the information, the program prediction process is triggered, and the system automatically performs network extraction, construction and calculation, and outputs the probability of each complaint generated by the batch of products. Through the secondary development of MES, the judgment of quality control point abnormalities can be realized in the MES system, and the abnormal information can be structured and reported to the complaint prediction system, triggering the complaint prediction process, and finally visually publishing the quality problems and prediction results to the information board for decision-making. This process realizes the full process automatic mapping from quality abnormalities to quality costs.

[0105] The system can be integrated with the enterprise knowledge graph system to dynamically extract the knowledge network of quality control points to adapt to the enterprise process improvement information, reduce the risk of data inconsistency, and reduce the manual configuration of system processes. It can also be integrated with the enterprise-level parallel computing engine to achieve rapid implementation of neural network calculations, thereby achieving quasi-real-time execution of complaint predictions, so that enterprises can make agile decisions after discovering quality control anomalies.

[0106] Beneficial effects:

[0107] This application effectively solves the problem of quality information waste in liquid milk production workshops and improves data application capabilities. This application designs a fusion processing method for quality inspection information and quality complaint information, as well as a training method for the fusion processed data set, which can effectively utilize the information of the production process technology management and after-sales service management departments, and mine the inherent correlation between information through artificial intelligence algorithms. This technology breaks through the barriers of intercommunication of quality information records between departments, giving the original statistical data intelligent prediction capabilities.

[0108] This application effectively solves the problem of cumbersome information records for workers in the liquid milk workshop and improves digital office capabilities. This application designs a human-computer interaction interface for the liquid milk quality problem complaint prediction function, and based on this, develops a liquid milk product complaint prediction system, allowing workers to intuitively operate the graphical interface, enter quality problem inspection information, and collect information with one click to generate quality problem record charts and complaint type prediction results. This technology replaces the paper-based work method of recording quality inspection problems, improves workers' office efficiency and convenience, and improves information operation efficiency.

[0109] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store video tag processing data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for predicting complaints about liquid milk product quality problems is implemented.

[0110] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0111] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0112] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0113] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0114] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0115] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0116] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0117] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0118] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for predicting complaints about liquid milk product quality problems, characterized in that: The liquid milk product quality problem complaint prediction method comprises: Acquire quality control point list data; the quality control point list data includes: corresponding control parameters of each sub-process in the process link during the liquid milk production process; the process link includes: raw milk pre-treatment, milk filling and milk packaging; the control parameters include: buffer chain lubrication level, main compressed air pressure and straw heating temperature; The quality control point list data is input into the liquid milk product complaint prediction model, and the predicted complaint data is output; the predicted complaint data includes: complaint type and complaint probability; the liquid milk product complaint prediction model is obtained by training a constructed fully connected neural network based on the historical complaint causal relationship data set of each process link; the liquid milk product complaint prediction model includes: a raw milk pre-processing complaint prediction sub-model, a milk filling complaint prediction sub-model and a milk packaging complaint prediction sub-model; the fully connected neural network includes an input layer, a hidden layer and an output layer connected in sequence; the historical complaint causal relationship data set is a coded sequence data determined by matching historical complaint information, historical inspection problem information and historical liquid milk product quality control point list data according to the set control point parameter setting method.

2. The method for predicting complaints about liquid milk product quality problems according to claim 1, characterized in that: The method for determining the liquid milk product complaint prediction model specifically includes: Obtain historical complaint causal relationship dataset; Dividing the historical complaint causal relationship dataset into a training set and a test set according to a set ratio; Build a fully connected neural network; The training set is input into the hidden layer through the input layer, and the weight parameters of the hidden layer are trained and updated by using the stochastic gradient descent method and the momentum algorithm with the objective of minimizing the objective function, so as to obtain the trained hidden layer; the objective function is determined according to the deviation between the output value of the hidden layer and the actual value of the historical complaint information; The test set is used to adjust the parameters of the trained hidden layer to obtain an adjusted hidden layer; the parameters include: learning rate and momentum parameter; The output layer is used to use the tanh function to perform nonlinear relationship fitting, and uses the loss function and the softmax function to determine the historical complaint data according to the output value of the adjusted hidden layer; Determining a trained fully connected neural network according to the input layer, the adjusted hidden layer and the output layer; The trained fully connected neural network is determined as the liquid milk product complaint prediction model.

3. The method for predicting complaints about liquid milk product quality problems according to claim 2, characterized in that: The acquisition of the historical complaint causal relationship dataset specifically includes: Obtain historical inspection problem information, historical complaint information, and historical liquid milk product quality control point list data; According to the historical complaint information and the historical inspection problem information, information matching is performed to obtain first matching relationship information; Perform information matching according to the first matching relationship information, the historical inspection problem information, and the historical liquid milk product quality control point list data to obtain second matching relationship information; According to the historical complaint information and the second matching relationship information, the coding sequence data is determined in accordance with the setting control point parameter setting method; The historical complaint causal relationship dataset is determined based on the coding sequence data.

4. The method for predicting complaints about liquid milk product quality problems according to claim 1, characterized in that: The liquid milk product quality problem complaint prediction method further includes: Comparing the complaint prediction data with a preset threshold value to obtain a comparison result; the preset threshold value includes: a first threshold value and a second threshold value; the first threshold value is less than the second threshold value; If the predicted data is greater than the first threshold but less than the second threshold, a prompt message is issued; If the predicted data is greater than or equal to the second threshold, a warning message is issued; the warning message is used to prompt the staff to conduct product testing.

5. The method for predicting complaints about liquid milk product quality problems according to claim 2, characterized in that: The loss function used is the cross entropy loss function; The expression of the loss function is: Where L is the loss function; N is the total number of data in the historical complaint causal relationship dataset; i is the sequence number; j is the label of the complaint type in the complaint prediction data; c ij is a constant value indicating whether the output value of the i-th group of data is really a complaint of the j-th type; p ij For complaint prediction data, the probability of complaint; p ij It is obtained by processing with the softmax function, and the calculation formula is: y ij is the output value of the adjusted hidden layer.

6. The method for predicting complaints about liquid milk product quality problems according to claim 1, characterized in that: The liquid milk product quality problem complaint prediction method further includes: A knowledge graph method is used to perform visualization based on the predicted complaint data.

7. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for predicting complaints about quality problems of liquid milk products according to any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting complaints about quality issues of liquid milk products according to any one of claims 1 to 6 is implemented.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for predicting complaints about quality issues of liquid milk products according to any one of claims 1 to 6 is implemented.

Images