User satisfaction analysis processing method and device

Abstract

The invention discloses a user satisfaction analysis processing method and device, and the method comprises the steps: obtaining user behavior data on a search result page in real time, carrying out the real-time data processing of the user behavior data, obtaining summary information, carrying out the satisfaction evaluation through a pre-trained machine learning model and the summary information, and obtaining the user satisfaction. Through a dynamic search result optimization mechanism, the user satisfaction is optimized, an optimized search result is obtained and displayed, a machine learning model is trained and adjusted according to the optimized search result, a closed-loop feedback system is formed, and continuous optimization of the search result is ensured.

Description

Technical Field

[0001] The present application relates to the technical field of data analysis, and more specifically, to a method and device for analyzing and processing user satisfaction. Background Art

[0002] In modern search technology, a key challenge is to improve the relevance of search results and user satisfaction. Traditional search engines usually provide search results based on keyword matching and historical data analysis, but this method often fails to accurately capture users' real-time needs and preferences, resulting in search results that do not match user expectations. This inaccuracy not only affects the user experience, but may also lead to user churn.

[0003] In terms of improving search engine accuracy and user satisfaction, the solutions currently commonly used in the industry mainly include personalized search algorithms, content-based recommendations, click-through rate prediction models, etc. However, personalized search algorithms, content-based recommendations, and click-through rate prediction models have the following shortcomings: Personalized search algorithms mainly rely on historical data and cannot accurately capture users' real-time search intentions and changes. Content-based recommendations have limitations in understanding users' specific needs, especially when user queries are ambiguous or ambiguous, the system may not be able to provide accurate results. Although the click-through rate prediction model can predict the popularity of a search result, it may not accurately reflect the user's true satisfaction with the search result.

[0004] Therefore, how to accurately capture the real-time needs of users and provide a personalized search experience is an urgent problem to be solved in this application. Summary of the invention

[0005] In view of this, the present application discloses a user satisfaction analysis and processing method and device, which aims to improve the search engine usage efficiency and satisfaction of user satisfaction, thereby accurately capturing the real-time needs of users and providing a personalized search experience.

[0006] In order to achieve the above purpose, the disclosed technical solution is as follows:

[0007] The first aspect of the present application discloses a method for analyzing and processing user satisfaction, the method comprising:

[0008] Get user behavior data on search results pages in real time;

[0009] Performing real-time data processing on the user behavior data to obtain summary information;

[0010] Performing satisfaction evaluation using a pre-trained machine learning model and the summary information to obtain user satisfaction;

[0011] By means of a dynamic search result optimization mechanism, the user satisfaction is optimized, and optimized search results are obtained and displayed;

[0012] The optimized search results are used to train and adjust the machine learning model to form a closed-loop feedback system to ensure continuous optimization of search results.

[0013] Preferably, the real-time acquisition of user behavior data on the search result page includes:

[0014] The user behavior is tracked in real time to obtain the user behavior data on the search result page; wherein the user behavior data at least includes the number of clicks, dwell time, scroll depth, return frequency and sequence pattern analysis.

[0015] Preferably, the real-time data processing at least includes data cleaning, feature extraction, data standardization, and real-time aggregation and summarization. The real-time data processing of the user behavior data to obtain summary information includes:

[0016] The user behavior data is subjected to the data cleaning, feature extraction, data standardization and machine learning analysis to obtain summary information.

[0017] Preferably, the training process of the machine learning model includes:

[0018] Obtain training data on click counts, page dwell time, and page scroll depth;

[0019] Constructing a neural network; wherein the neural network at least includes an input layer, a hidden layer and an output layer;

[0020] In the process of training the neural network, the training data of the number of clicks, the training data of the page dwell time and the training data of the page scroll depth are used to train the neural network to calculate the predicted value of the output layer of the neural network;

[0021] Use the cross entropy loss function to calculate the loss of the neural network and obtain the calculated loss function;

[0022] By calculating the loss function, the predicted value and the gradient of the network parameters, the weight and bias of the neural network are updated, and the step of training the neural network with the training data of the number of clicks, the training data of the page dwell time and the training data of the page scroll depth is returned to be executed in the process of training the neural network to calculate the predicted value of the output layer of the neural network, so as to repeat multiple rounds of training iterations;

[0023] In the process of repeating multiple rounds of training iterations, the network parameters of the neural network are continuously adjusted, and the accuracy and generalization ability of the machine learning model are evaluated through the validation set to complete the training process of the machine learning model.

[0024] Preferably, the satisfaction evaluation is performed using the pre-trained machine learning model and the summary information to obtain user satisfaction, including:

[0025] Preprocessing the summary information;

[0026] Define the structure of the neural network in the pre-trained machine learning model to determine the number of network layers of the neural network and the number of neurons in each layer corresponding to the number of network layers;

[0027] Randomly initialize the weights and biases of the defined neural network;

[0028] By initializing the weights and biases of the neural network, the preprocessed summary information is weighted and activated to obtain the output value;

[0029] The difference between the output value and the actual satisfaction label is evaluated by the loss function to obtain the user satisfaction.

[0030] Preferably, it also includes:

[0031] Calculate the gradient of the network parameters of the neural network and update the weights and bias of the neural network using gradient descent.

[0032] Preferably, the user satisfaction is optimized through a dynamic search result optimization mechanism to obtain and display optimized search results, including:

[0033] Using the user satisfaction as a new ranking factor, and combining the new ranking factor with the traditional ranking factor to obtain a combined ranking factor;

[0034] The search ranking algorithm is adjusted and optimized according to the combined ranking factors, so as to dynamically adjust the weight and ranking of each search result according to the user satisfaction score, and obtain and display optimized search results.

[0035] The second aspect of the present application discloses a user satisfaction analysis and processing device, the device comprising:

[0036] A real-time acquisition unit, used to acquire user behavior data on the search result page in real time;

[0037] A real-time data processing unit, used to perform real-time data processing on the user behavior data to obtain summary information;

[0038] An evaluation unit, used to perform satisfaction evaluation through a pre-trained machine learning model and the summary information to obtain user satisfaction;

[0039] An optimization unit, used to optimize the user satisfaction through a dynamic search result optimization mechanism, obtain optimized search results and display them;

[0040] A training adjustment unit is used to train and adjust the machine learning model based on the optimized search results to form a closed-loop feedback system to ensure continuous optimization of the search results.

[0041] Preferably, the real-time acquisition unit is specifically used for:

[0042] The user behavior is tracked in real time to obtain the user behavior data on the search result page; wherein the user behavior data at least includes the number of clicks, dwell time, scroll depth, return frequency and sequence pattern analysis.

[0043] Preferably, the real-time data processing at least includes data cleaning, feature extraction, data standardization, and real-time aggregation and summarization, and the real-time data processing unit is specifically used for:

[0044] The user behavior data is subjected to the data cleaning, feature extraction, data standardization and machine learning analysis to obtain summary information.

[0045] It can be seen from the above technical solution that the present application discloses a user satisfaction analysis and processing method and device, which obtains user behavior data on the search results page in real time, performs real-time data processing on the user behavior data to obtain summary information, conducts satisfaction evaluation through pre-trained machine learning models and summary information to obtain user satisfaction, optimizes user satisfaction through a dynamic search result optimization mechanism, obtains and displays optimized search results, trains and adjusts the machine learning model with the optimized search results, and forms a closed-loop feedback system to ensure continuous optimization of search results.

[0046] Through the above solution, by tracking and analyzing the user's interactive behavior on the search results, such as click-through rate, page dwell time, etc., the user's satisfaction with the search results is evaluated in real time, a personalized search experience is provided, and based on the feedback from the evaluation of the user's satisfaction with the search results, the search algorithm is adjusted and optimized. That is, through the continuous learning and adjustment of the machine learning model, the system can accurately capture the user's current needs and preferences, so that the final search results can better match the user's actual needs, improve the user's satisfaction, and improve the search engine usage efficiency and satisfaction, thereby accurately capturing the user's real-time needs and providing a personalized search experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] 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 or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0048] Figure 1 A flowchart of a user satisfaction analysis and processing method disclosed in an embodiment of the present application;

[0049] Figure 2 A schematic diagram of the structure of a neural network disclosed in an embodiment of the present application;

[0050] Figure 3 A schematic diagram of a closed-loop feedback system disclosed in an embodiment of the present application;

[0051] Figure 4 A flowchart of another method for analyzing and processing user satisfaction disclosed in an embodiment of the present application;

[0052] Figure 5 A schematic diagram of the structure of a user satisfaction analysis and processing device is disclosed for an embodiment of the present application;

[0053] Figure 6 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of the present application. DETAILED DESCRIPTION

[0054] 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.

[0055] In this application, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0056] As can be seen from the background technology, in terms of improving search engine accuracy and user satisfaction, the solutions currently commonly used in the industry mainly include personalized search algorithms, content-based recommendations, click-through rate prediction models, etc. However, personalized search algorithms, content-based recommendations, and click-through rate prediction models have the following shortcomings: personalized search algorithms mainly rely on historical data and cannot accurately capture users' real-time search intentions and changes. Content-based recommendation methods have limitations in understanding users' specific needs, especially when user queries are vague or ambiguous, the system may not be able to provide accurate results. Although the click-through rate prediction model can predict the popularity of a certain search result, it may not accurately reflect the user's true satisfaction with the search result. Therefore, how to accurately capture users' real-time needs and provide a personalized search experience is a problem that this application urgently needs to solve.

[0057] In order to solve the above problems, the present application discloses a method and device for analyzing and processing user satisfaction, which can evaluate the user's satisfaction with the search results in real time by real-time tracking and analyzing the user behavior data of the user's interactive behavior with the search results, such as click-through rate, page dwell time, etc., and provide a personalized search experience. Based on the feedback of evaluating the user's satisfaction with the search results, the search algorithm is adjusted and optimized, that is, through the continuous learning and adjustment of the machine learning model, the system can accurately capture the user's current needs and preferences, so that the final search results can better match the user's actual needs, improve the search engine usage efficiency and satisfaction of the user, and thus accurately capture the user's real-time needs and provide a personalized search experience. The specific implementation method is specifically described through the following embodiments.

[0058] This application optimizes search results based on real-time user feedback. By tracking and analyzing user interaction with search results (such as click-through rate, page dwell time, etc.), it evaluates user satisfaction with search results in real time, and adjusts and optimizes search algorithms based on these feedbacks. This method can more accurately understand users' real-time needs and preferences, thereby providing more relevant and satisfactory search results. The basic scheme is as follows:

[0059] (1) Behavior tracking module, which captures and records in real time the user's click behavior on search results, page browsing time, return or re-search behavior, etc.

[0060] (2) Data analysis engine, which analyzes the collected behavioral data to determine which search results are considered useful and satisfactory by users and which are not.

[0061] (3) Satisfaction evaluation algorithm: Based on user behavior data, it uses machine learning algorithms (such as logistic regression, decision tree, etc.) to evaluate the user's satisfaction with each search result.

[0062] Among them, logistic regression is a widely used machine learning algorithm for predicting the probability of a result, such as a binary classification problem. In this application, it is used to predict whether the user's satisfaction with the search results is high or low.

[0063] Decision tree is a machine learning method used for classification and regression. It predicts the output by creating a decision path. It is used to identify factors that lead to satisfactory or unsatisfactory search results for users.

[0064] (4) Search algorithm optimization: Based on the satisfaction evaluation results, adjust the search engine’s ranking algorithm to prioritize results that are more likely to meet user needs.

[0065] (5) Real-time update mechanism to ensure that the system can respond to user feedback in real time, quickly adjust search results, and provide real-time and relevant information.

[0066] Through the above method, this application aims to significantly improve the accuracy and user satisfaction of search engines, thereby improving the user experience and bringing higher user retention and satisfaction to search engine providers.

[0067] refer to Figure 1 As shown, a user satisfaction analysis and processing method disclosed in an embodiment of the present application is provided, and the user satisfaction analysis and processing method mainly includes the following steps:

[0068] S101: Obtain user behavior data on the search result page in real time.

[0069] In S101, user behavior tracking is performed in real time to obtain user behavior data on the search result page; wherein the user behavior data includes but is not limited to the number of clicks, dwell time, scroll depth, return frequency and sequence pattern analysis.

[0070] Behavior tracking refers to monitoring and recording user activities on search engine results pages, such as clicks on links, page browsing time, scrolling behavior, etc. By analyzing these user behavior data through the Data Analysis Engine, we can understand the user's satisfaction and preference for search results.

[0071] The data analysis engine is a set of algorithms and programs that process and analyze the collected user behavior data. It aims to extract useful information from user behavior, such as which search results are more popular and which fail to meet user needs, so as to determine which search results are considered useful and satisfactory by users, and which search results are considered useless and unsatisfactory by users.

[0072] Real-time capture and recording of user behavior data corresponding to various interactive behaviors of users on the search results page. The source of user behavior data includes but is not limited to the client-integrated software development kit (SDK) collection plug-in for user clicks, browsing, exits and other behavior data on the page, as well as the system logs of key nodes such as search word input and search output, and batch data.

[0073] Among them, the number of clicks refers to the number of times users click on the search results. The number of clicks is used to evaluate which search results are most popular or most in line with the user's search intent.

[0074] Dwell time refers to how long a user stays on a specific search results page, where a long stay may indicate that the user is interested in or satisfied with the content.

[0075] Scroll depth refers to the distance or proportion that a user scrolls on a page. It may mean that the content that the user is interested in does not appear on the first screen, and the result sorting needs to be optimized.

[0076] Return frequency refers to how often users return to the search results page. Frequent returns may indicate that users did not find the information they were looking for in the search results.

[0077] Sequential pattern analysis refers to the order and pattern of users' jumps between different search results. Sequential pattern analysis represents the user's decision path and preferences during the search process.

[0078] S102: Perform real-time data processing on user behavior data to obtain summary information.

[0079] It should be noted that real-time data processing is one of the core links in the present invention, which ensures that the collected user behavior data can be quickly and accurately analyzed and applied to search result optimization.

[0080] This application adopts a data stream processing framework such as a big data platform (Zero DiscoverZero Hack, ZDH) (including distributed parallel stream processing system (flink) real-time processing, open source stream processing platform (Kafka) real-time transmission) + columnar database management system (clickhouse), which can process large-scale, continuous data streams in real time. At the same time, the ZDH framework supports distributed data processing to ensure that the system can handle a large amount of concurrent user behavior data. In addition, this application uses efficient data structures and algorithms to minimize data processing delays, ensuring that user behavior data can be analyzed as quickly as possible and used to drive the optimization of search results.

[0081] Real-time data processing includes at least data cleaning, feature extraction, data standardization, and real-time aggregation and summarization.

[0082] Data cleaning is used to remove invalid, erroneous or incomplete data records, thereby improving data quality and ensuring the accuracy of subsequent analysis.

[0083] Feature extraction is used to extract key features from raw data, such as click-through rate, dwell time, scrolling behavior, etc., thereby converting the raw data into an input format that can be used by machine learning models.

[0084] Data standardization is used to convert data to a uniform format and range, such as by scaling or transforming the data, thereby eliminating biases between different data types and magnitudes and improving the performance of machine learning models.

[0085] Real-time aggregation and summarization are used to aggregate data in real time and generate summary information, such as average dwell time, to provide a quick data overview and support real-time decision making and analysis.

[0086] In S102, data cleaning, feature extraction, data standardization and machine learning analysis are performed on the user behavior data to obtain summary information.

[0087] The summary information includes but is not limited to the average residence time, which provides a quick data overview and supports real-time decision-making and analysis.

[0088] S103: Perform satisfaction evaluation through the pre-trained machine learning model and summary information to obtain user satisfaction.

[0089] In S103, the satisfaction evaluation module uses a neural network machine learning algorithm to accurately evaluate the user's satisfaction with the search results.

[0090] Through neural networks, users’ satisfaction with search results is evaluated based on their behavior on the search results page.

[0091] A neural network is an algorithmic framework that simulates the working principles of neurons in the human brain and is widely used in areas such as pattern recognition, data classification, and predictive analysis.

[0092] The specific structure of the neural network is as follows Figure 2 shown.

[0093] Figure 2 In the neural network, there are input layers, hidden layers and output layers.

[0094] The input layer is used to receive the raw data from external input. Each input layer node represents a feature in the data set. In user satisfaction evaluation, the input layer node can represent features such as the number of user clicks and page dwell time.

[0095] One or more hidden layers in a neural network are responsible for complex transformations and processing of input data. Each hidden layer consists of multiple neurons, which can capture and learn nonlinear relationships in the data. The number of neurons and the number of layers in the hidden layer can be adjusted, affecting the complexity and learning ability of the model.

[0096] The output layer is responsible for converting the processing results of the hidden layer into the final output, such as classification or prediction results. In user satisfaction evaluation, the output layer may have only one neuron, indicating whether the user is satisfied with the search results.

[0097] The working mode of neural network includes signal forward propagation (ForwardPropagation) and error backpropagation (Backpropagation). The specific working mode of neural network is explained as follows:

[0098] (1) Signal forward propagation: Each neuron receives input signals from neurons in the previous layer; the input signals are weighted and added with a bias value; the weighted sum is transformed by an activation function to introduce nonlinearity, enabling the neural network to learn and simulate complex data relationships; the output of the activation function becomes the input of the next layer of neurons.

[0099] Among them, the activation function introduces nonlinearity so that the neural network can handle nonlinear problems. Common activation functions include but are not limited to Sigmoid, ReLU (Rectified Linear Unit), Tanh (Hyperbolic Tangent), etc. Weights and Biases, weights are used to control the influence of the previous layer of neuron signals on the current neuron output. Biases are used to adjust the output of neurons to make them closer to the desired output.

[0100] (2) Error backpropagation: During training, the difference between the output of the neural network and the expected output (label) is used to adjust the weights and biases; the error is propagated backward through the network, and the weights and biases of each layer are adjusted according to their contribution to the final error.

[0101] Through these principles, neural networks are able to learn complex data patterns and effectively predict and classify new input data. In this application, neural networks are used to learn from user behavior data to accurately evaluate user satisfaction with search results.

[0102] The specific process of evaluating satisfaction through a pre-trained machine learning model and summary information is shown in A1-A5.

[0103] A1: Preprocess the summary information.

[0104] In A1, user behavior data (such as number of clicks, dwell time, etc.) is used as feature input and preprocessed.

[0105] Preprocessing includes but is not limited to standardization (ensuring that the input data is of the same scale), missing value processing, etc.

[0106] A2: Define the structure of the neural network in the pre-trained machine learning model to determine the number of network layers of the neural network and the number of neurons in each layer corresponding to the number of network layers.

[0107] Among them, the neural network structure is defined, the number of network layers and the number of neurons in each layer are determined, and the activation function (such as ReLU or Sigmoid) is selected to introduce nonlinearity in the network.

[0108] A3: Randomly initialize the weights and biases of the defined neural network.

[0109] A4: After initializing the weights and biases of the neural network, the preprocessed summary information is weighted and activated to obtain the output value.

[0110] In A4, the neural network after initializing weights and biases is used to perform weighted and activation function calculations on the summary information through forward propagation calculations. That is, for each training sample, the neural network calculates the final output through the weighted sum between layers and the activation function, and the expression of the output value is shown in formula (1).

[0111] y=f(W·x+b)(1)

[0112] Among them, y is the output value; f is the activation function; W is the weight matrix; x is the input vector; b is the bias.

[0113] A5: The difference between the output value and the actual satisfaction label is evaluated through the loss function to obtain the user satisfaction.

[0114] In A5, the loss is calculated to evaluate the difference between the output value and the actual satisfaction label. Specifically, a loss function (such as cross entropy loss) is used to evaluate the difference between the network output and the actual satisfaction label.

[0115] Through back propagation and weight update, the gradient of the loss function with respect to the network parameters of the neural network is calculated, and the weights and biases of the neural network are updated using gradient descent.

[0116] The derivation of the machine learning algorithm of the neural network includes signal forward propagation, loss function, back propagation and weight update. The derivation of the machine learning algorithm of the specific neural network is as follows:

[0117] Signal forward propagation:

[0118] The output calculation formula of each layer is shown in formula (2):

[0119] O l =f(W l ·O l-1 +b l )(2)

[0120] Among them, O l is the input of the lth layer, which is the output weighted function of the l-1th layer; W l is the weight of the lth layer; b l is the bias of the lth layer; f is the activation function.

[0121] Loss function:

[0122] Assuming the cross entropy loss function is used, for the binary classification problem, the loss function is calculated as shown in formula (3).

[0123]

[0124] in, is the loss function; y is the true label; is the predicted label.

[0125] Backward Propagation:

[0126] The purpose of back propagation is to calculate the gradient of the loss function with respect to each parameter. The gradient calculation formula is shown in formula (4).

[0127]

[0128] in, is the gradient of the loss function with respect to the output layer weights; is the gradient of the loss function with respect to the weighted input of the output layer; Gradient of the weighted input to the output layer with respect to the output layer weights.

[0129] Gradients are back-propagated to each layer via the chain rule.

[0130] Weight update:

[0131] The weight update formula is shown in formula (5).

[0132]

[0133] Among them, α is the learning rate, which is used to control the update step size; W l is the weight of the lth layer.

[0134] Through the derivation process of the machine learning algorithm of the neural network described above, the neural network is able to learn from user behavior data and accurately evaluate user satisfaction with search results.

[0135] The specific training process of the machine learning model is shown in B1-B5.

[0136] B1: Obtain training data on the number of clicks, the time spent on a page, and the depth of page scrolling.

[0137] In order to evaluate users' satisfaction with search results based on their behavior in search engines, this solution collects training data on user behavior, including training data on clicks, training data on page dwell time, and training data on page scroll depth.

[0138] 1) Data preprocessing:

[0139] Use the above training data to perform data preprocessing. Assume that a large amount of user behavior data has been collected and converted into a standardized format. For example, the dwell time and scroll depth may be standardized to a range of 0 to 1.

[0140] Use the above training data to train a neural network to predict whether the user is satisfied with the search result (yes / no).

[0141] B2: Construct a neural network; wherein the neural network at least includes an input layer, a hidden layer and an output layer.

[0142] For example, a simple three-layer neural network is constructed:

[0143] Input layer: 3 neurons, corresponding to 3 features (number of clicks, dwell time, scroll depth).

[0144] Hidden layer: 5 neurons, using ReLU activation function.

[0145] Output layer: 1 neuron, using Sigmoid activation function to predict user satisfaction.

[0146] B3: In the process of training the neural network, the training data of the number of clicks, the training data of the page dwell time and the training data of the page scroll depth are used to train the neural network to calculate the predicted value of the output layer of the neural network.

[0147] Training process:

[0148] Forward propagation: For a given training sample (such as click count = 2, dwell time = 0.5, scroll depth = 0.8), the neural network calculates the predicted value of the output layer.

[0149] B4: Use the cross entropy loss function to calculate the loss of the neural network and obtain the calculated loss function.

[0150] Loss calculation: Use the cross entropy loss function and assume that the true user satisfaction label is "1" (satisfied) to calculate the loss.

[0151] Backpropagation: Update weights and biases by calculating the gradient of the loss function with respect to the network parameters.

[0152] B5: By calculating the gradient of the loss function, the predicted value and the network parameters, the weights and biases of the neural network are updated, and the training data of the number of clicks, the page dwell time and the page scroll depth are used to train the neural network to calculate the predicted value of the output layer of the neural network, so as to repeat multiple rounds of training iterations.

[0153] Weight update:

[0154] Suppose that in one training iteration, the network's output prediction is 0.6, while the true label is 1. The gradient of the loss function is calculated and used to update the network's weights and biases.

[0155] Evaluation and Optimization:

[0156] Repeat multiple rounds of training iterations, continuously adjust the network parameters, evaluate the accuracy and generalization ability of the model through the validation set, and then continue to optimize the model parameters based on the evaluation results.

[0157] B6: In the process of repeating multiple rounds of training iterations, the network parameters of the neural network are continuously adjusted, and the accuracy and generalization ability of the machine learning model are evaluated through the validation set to complete the training process of the machine learning model.

[0158] Through the above process, the neural network model can learn how to accurately predict users’ satisfaction with search results based on their clicks, dwell time, and scroll depth. This method can help us better understand users’ needs and preferences, thereby providing more personalized and accurate search results.

[0159] S104: Optimize user satisfaction through a dynamic search result optimization mechanism, obtain optimized search results and display them.

[0160] Among them, the dynamic search result optimization mechanism, that is, adjusting and optimizing the search algorithm according to real-time user satisfaction evaluation, focuses on taking the user's actual feedback as the key factor in optimizing the search results and re-sorting the search results, that is, adjusting the ranking order of the search results based on the satisfaction score so that the results with higher scores are displayed closer.

[0161] Specifically, user satisfaction is optimized through a dynamic search result optimization mechanism, and the process of obtaining and displaying optimized search results is shown in C1-C2.

[0162] Adjust the search ranking algorithm:

[0163] C1: Use user satisfaction as a new ranking factor and combine the new ranking factor with traditional ranking factors (such as keyword matching and page authority) to obtain a combined ranking factor.

[0164] C1 is the ranking factor update.

[0165] C2: Adjust and optimize the search ranking algorithm based on the combined ranking factors to dynamically adjust the weight and ranking of each search result based on the user satisfaction score, obtain optimized search results and display them.

[0166] The algorithm logic is adjusted in C2 so that the weight and ranking of each search result can be dynamically adjusted according to the user satisfaction score.

[0167] The machine learning model is trained through feedback data, and the ranking order of search results is dynamically adjusted according to user satisfaction scores (so that results with higher scores are displayed closer to the front), to obtain and display optimized search results.

[0168] The training process of the feedback data training machine learning model is as follows:

[0169] Dataset construction: collect user behavior data and corresponding satisfaction scores to build a training dataset for machine learning;

[0170] Model training: Use machine learning algorithms (such as neural networks and gradient boosting trees) to train data and predict the potential satisfaction of unknown search results by learning the relationship between user behavior and satisfaction.

[0171] Model iterative optimization: regularly update the model based on newly collected data to capture changes in user preferences and emerging search trends, and achieve continuous learning and optimization of the model.

[0172] Optimizing search results includes personalized display and interactive optimization, as follows:

[0173] Personalized display: adjust the display of search results based on the user's unique preferences and historical behavior, such as customization of layout and summary information;

[0174] Interaction optimization: optimize the way users interact with search results, such as adding a quick feedback mechanism so that users can express their satisfaction or dissatisfaction with the results more directly.

[0175] Through the above optimization mechanism and implementation steps, the search engine of the present application can respond to the real-time needs of users more flexibly and accurately, enabling the search engine to adapt to changing user behaviors and preferences, thereby significantly improving the relevance of search results and overall satisfaction.

[0176] S105: The machine learning model will be trained and adjusted to optimize the search results, forming a closed-loop feedback system to ensure continuous optimization of the search results.

[0177] The system integrates data collection, processing, machine learning evaluation and search optimization modules, and continuously learns and adjusts based on new user feedback data to form a closed-loop feedback system to ensure continuous optimization of search results. Figure 3 shown. Figure 3 A schematic diagram of a closed-loop feedback system is shown.

[0178] Figure 3 Collect user behavior data (number of clicks, browsing time, etc.);

[0179] Preprocess user behavior data (cleaning, feature extraction);

[0180] Train machine learning models (Logistic Regression, neural networks, etc.);

[0181] Evaluate user satisfaction through trained machines (based on model analysis);

[0182] Update ranking factors (adjust ranking algorithm);

[0183] Optimize user satisfaction according to the adjusted ranking algorithm. During the optimization process, re-rank the search results according to the adjusted ranking algorithm (based on the new ranking factors), obtain the optimized search structure and display it (adjust the display method);

[0184] Collect new user feedback, that is, new user behavior data, for model optimization, and return to the step of training the machine learning model to form a closed-loop feedback system to ensure continuous optimization of search results.

[0185] In order to facilitate the understanding of the process of user satisfaction analysis, combined with Figure 4 Let's take an example:

[0186] Figure 4 In the process, user behaviors are tracked (such as clicks, dwell time, scrolling, etc.) to obtain user behavior data;

[0187] Perform real-time data processing on user behavior data (using technologies such as Apache Flink and Kafka), including data cleaning (removing invalid / incomplete data), feature extraction (extracting key features from the data), data standardization (standardizing data format and scale), and machine learning analysis (performing machine learning analysis by applying algorithms such as logistic regression);

[0188] Conduct satisfaction evaluation on user behavior data after real-time data processing to evaluate user satisfaction with search results;

[0189] Optimize search results, obtain and display optimized search results, and adjust search algorithms based on the optimized search results;

[0190] Optimize search results through search algorithm feedback and integrate the optimized search results into the system to form a closed-loop feedback system to ensure continuous optimization of search results.

[0191] Advantages of this application:

[0192] (1) Improve the relevance and accuracy of search results. Collect and analyze users’ interactive behaviors on search results in real time. The system can accurately capture users’ current needs and preferences. Through continuous learning and adjustment of machine learning models, search results can better match users’ actual needs, thereby improving search engine usage efficiency and satisfaction.

[0193] (2) Enhance user experience. When search results more accurately reflect user needs, they can provide a more satisfying and enjoyable search experience, thereby enhancing user loyalty and the attractiveness of the platform.

[0194] (3) Dynamically respond to changes in user needs, process and analyze user behavior data in real time, and respond quickly to changes in user needs. Continuous optimization of search algorithms ensures that search results are adjusted as user preferences change, and can adapt to dynamic changes in user needs, always providing the most relevant search results.

[0195] (4) Reduced reliance on explicit user feedback: The present invention reduces the need for explicit user feedback by analyzing users’ implicit behaviors (such as clicks and browsing time) to evaluate their satisfaction with search results. Even when users are unwilling or unable to provide direct feedback, the search engine can continue to improve.

[0196] (5) Data-driven decision support. By collecting a large amount of user interaction data, the present invention provides search engine providers with rich data support for optimizing decision-making and strategy formulation, promoting improved search algorithms and user interface design, and making the operation of search engines more efficient.

[0197] The core technologies of this application include behavioral data collection, real-time data processing, machine learning-driven satisfaction evaluation, and dynamic search result optimization. This application has significant advantages in improving search accuracy, user experience, adaptability, user feedback collection, and data-driven decision-making, which improves the overall performance and competitiveness of search engines and meets users' needs for efficient, accurate, and personalized search services.

[0198] By monitoring user behavior data in real time and using machine learning for data analysis, this application can more accurately evaluate user satisfaction and dynamically adjust search algorithms to provide more relevant results, addressing the shortcomings of traditional technologies in dynamically capturing user needs and real-time feedback.

[0199] The core of this application is to improve the accuracy and user satisfaction of search engines by accurately capturing user behaviors and preferences. The following are the key points and protection points of this application:

[0200] (1) User Satisfaction Assessment Algorithm: Use machine learning algorithms (such as logistic regression, decision tree, neural network) to analyze user behavior data and evaluate user satisfaction with search results. To protect the point, the specific design of the machine learning model, including algorithm selection, model training method, and how to extract effective features from user behavior data.

[0201] The user satisfaction evaluation algorithm is a machine learning algorithm that is used to evaluate the user's satisfaction with specific search results. It uses the user's behavior data to predict their satisfaction with the search results, thereby optimizing the performance of the search engine.

[0202] (2) Dynamic optimization mechanism for search results: dynamically adjust the search algorithm based on the user satisfaction evaluation results to optimize the sorting and presentation of search results. To protect the point, the search algorithm optimization process includes how to adjust the search ranking and result display based on user satisfaction feedback.

[0203] Among them, search algorithm optimization refers to the adjustment and optimization of the underlying algorithm of the search engine to improve the relevance of search results and user satisfaction, ensuring that the search engine can provide more accurate and relevant search results based on real-time feedback from users.

[0204] (3) Real-time data processing and analysis framework, using real-time data processing technology to process and analyze user behavior data. To protect the point, the specific design of the real-time data processing architecture, including data flow management, data preprocessing and real-time analysis strategy.

[0205] (4) Feedback loop and continuous learning mechanism, implement a closed-loop feedback system so that the search engine can continuously learn and optimize itself based on the latest user feedback. To protect the point, the design and implementation of the feedback loop, including how to update the machine learning model and search algorithm based on new user data.

[0206] The beneficial effects of the embodiments of the present application are as follows: by real-time tracking and analyzing user behavior data of user interaction with search results, such as click-through rate, page dwell time, etc., user satisfaction with search results is evaluated in real time, a personalized search experience is provided, and based on feedback from evaluating user satisfaction with search results, the search algorithm is adjusted and optimized, that is, through continuous learning and adjustment of the machine learning model, the system can accurately capture the user's current needs and preferences, so that the final search results can better match the user's actual needs, thereby improving the search engine usage efficiency and satisfaction of user satisfaction, thereby accurately capturing the user's real-time needs and providing a personalized search experience.

[0207] Based on the above embodiments Figure 1 A user satisfaction analysis and processing method is disclosed, and the embodiment of the present application also discloses a user satisfaction analysis and processing device, the user satisfaction analysis and processing device is as follows Figure 5 As shown, the user satisfaction analysis and processing device includes:

[0208] A real-time acquisition unit 501 is used to acquire user behavior data on the search result page in real time;

[0209] A real-time data processing unit 502 is used to perform real-time data processing on user behavior data to obtain summary information;

[0210] An evaluation unit 503 is used to perform satisfaction evaluation through a pre-trained machine learning model and summary information to obtain user satisfaction;

[0211] An optimization unit 504 is used to optimize user satisfaction through a dynamic search result optimization mechanism, obtain optimized search results and display them;

[0212] The training adjustment unit 505 is used to train and adjust the machine learning model based on the optimized search results to form a closed-loop feedback system to ensure continuous optimization of the search results.

[0213] Furthermore, the real-time acquisition unit 501 is specifically used to track user behavior in real time to obtain user behavior data on the search result page; wherein the user behavior data at least includes the number of clicks, dwell time, scroll depth, return frequency and sequence pattern analysis.

[0214] Furthermore, real-time data processing includes at least data cleaning, feature extraction, data standardization, and real-time aggregation and summarization. The real-time data processing unit 502 is specifically used to perform data cleaning, feature extraction, data standardization and machine learning analysis on user behavior data to obtain summary information.

[0215] Furthermore, the evaluation unit 503 of the training process of the machine learning model includes:

[0216] The first acquisition module is used to acquire training data of click counts, page dwell time, and page scroll depth;

[0217] A construction module, used to construct a neural network; wherein the neural network at least includes an input layer, a hidden layer and an output layer;

[0218] A training module is used to train the neural network using the training data of the number of clicks, the training data of the page dwell time and the training data of the page scroll depth in the process of training the neural network, so as to calculate the predicted value of the output layer of the neural network;

[0219] A first calculation module is used to calculate the loss of the neural network using a cross entropy loss function to obtain a calculation loss function;

[0220] An update execution module is used to update the weights and biases of the neural network by calculating the gradient of the loss function, the predicted value and the network parameters, and return to execute the step of training the neural network with the training data of the number of clicks, the training data of the page dwell time and the training data of the page scroll depth in the process of training the neural network to calculate the predicted value of the output layer of the neural network, so as to repeat multiple rounds of training iterations;

[0221] The adjustment and evaluation module is used to continuously adjust the network parameters of the neural network during repeated rounds of training iterations, and evaluate the accuracy and generalization ability of the machine learning model through the validation set to complete the training process of the machine learning model.

[0222] Furthermore, the evaluation unit 503 includes:

[0223] A preprocessing module, used for preprocessing summary information;

[0224] A definition module is used to define the structure of the neural network in the pre-trained machine learning model to determine the number of network layers of the neural network and the number of neurons in each layer corresponding to the number of network layers;

[0225] Initialization module, used to randomly initialize the weights and biases of the defined neural network;

[0226] The second calculation module is used to perform weighted and activation function calculation on the preprocessed summary information through the neural network after initializing weights and biases to obtain an output value;

[0227] The evaluation module is used to evaluate the difference between the output value and the actual satisfaction label through the loss function to obtain the user satisfaction.

[0228] Furthermore, the user satisfaction analysis and processing device also includes:

[0229] The calculation update unit is used to calculate the gradient of the network parameters of the neural network and update the weights and bias of the neural network using gradient descent.

[0230] Furthermore, the optimization unit 504 includes:

[0231] A combining module, used for taking user satisfaction as a new ranking factor, and combining the new ranking factor with the traditional ranking factor to obtain a combined ranking factor;

[0232] The second acquisition module is used to adjust and optimize the search ranking algorithm according to the combined ranking factors, so as to dynamically adjust the weight and ranking of each search result according to the user satisfaction score, obtain optimized search results and display them.

[0233] The beneficial effects of the embodiments of the present application are as follows: by real-time tracking and analyzing user behavior data of user interaction with search results, such as click-through rate, page dwell time, etc., user satisfaction with search results is evaluated in real time, a personalized search experience is provided, and based on feedback from evaluating user satisfaction with search results, the search algorithm is adjusted and optimized, that is, through continuous learning and adjustment of the machine learning model, the system can accurately capture the user's current needs and preferences, so that the final search results can better match the user's actual needs, thereby improving the search engine usage efficiency and satisfaction of user satisfaction, thereby accurately capturing the user's real-time needs and providing a personalized search experience.

[0234] An embodiment of the present application further provides a storage medium, which includes stored instructions, wherein when the instructions are executed, the device where the storage medium is located is controlled to execute the user satisfaction analysis and processing method as described above.

[0235] The present application also provides an electronic device, whose structural diagram is as follows: Figure 6 As shown, it specifically includes a memory 601 and one or more instructions 602, wherein the one or more instructions 602 are stored in the memory 601 and are configured to be executed by one or more processors 603 to execute the one or more instructions 602 to perform the above-mentioned user satisfaction analysis processing method.

[0236] For the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0237] It should be noted that each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referred to each other. For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0238] The steps in the methods of the embodiments of the present application can be adjusted in order, combined, and deleted according to actual needs.

[0239] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0240] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0241] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for analyzing and processing user satisfaction, characterized in that: The method comprises: Get user behavior data on search results pages in real time; Performing real-time data processing on the user behavior data to obtain summary information; Performing satisfaction evaluation using a pre-trained machine learning model and the summary information to obtain user satisfaction; By means of a dynamic search result optimization mechanism, the user satisfaction is optimized, and optimized search results are obtained and displayed; The optimized search results are used to train and adjust the machine learning model to form a closed-loop feedback system to ensure continuous optimization of search results.

2. The method according to claim 1, characterized in that The real-time acquisition of user behavior data on the search results page includes: The user behavior is tracked in real time to obtain the user behavior data on the search result page; wherein the user behavior data at least includes the number of clicks, dwell time, scroll depth, return frequency and sequence pattern analysis.

3. The method according to claim 1, characterized in that The real-time data processing at least includes data cleaning, feature extraction, data standardization, and real-time aggregation and summarization. The real-time data processing of the user behavior data to obtain summary information includes: The user behavior data is subjected to the data cleaning, feature extraction, data standardization and machine learning analysis to obtain summary information.

4. The method according to claim 1, characterized in that: The training process of a machine learning model includes: Obtain training data on click counts, page dwell time, and page scroll depth; Constructing a neural network; wherein the neural network at least includes an input layer, a hidden layer and an output layer; In the process of training the neural network, the training data of the number of clicks, the training data of the page dwell time and the training data of the page scroll depth are used to train the neural network to calculate the predicted value of the output layer of the neural network; Use the cross entropy loss function to calculate the loss of the neural network and obtain the calculated loss function; By calculating the loss function, the predicted value and the gradient of the network parameters, the weight and bias of the neural network are updated, and the step of training the neural network with the training data of the number of clicks, the training data of the page dwell time and the training data of the page scroll depth is returned to be executed in the process of training the neural network to calculate the predicted value of the output layer of the neural network, so as to repeat multiple rounds of training iterations; In the process of repeating multiple rounds of training iterations, the network parameters of the neural network are continuously adjusted, and the accuracy and generalization ability of the machine learning model are evaluated through the validation set to complete the training process of the machine learning model.

5. The method according to claim 1, characterized in that: The satisfaction evaluation is performed by using the pre-trained machine learning model and the summary information to obtain the user satisfaction, including: Preprocessing the summary information; Define the structure of the neural network in the pre-trained machine learning model to determine the number of network layers of the neural network and the number of neurons in each layer corresponding to the number of network layers; Randomly initialize the weights and biases of the defined neural network; By initializing the weights and biases of the neural network, the preprocessed summary information is weighted and activated to obtain the output value; The difference between the output value and the actual satisfaction label is evaluated by the loss function to obtain the user satisfaction.

6. The method according to claim 5, characterized in that Also includes: Calculate the gradient of the network parameters of the neural network and update the weights and bias of the neural network using gradient descent.

7. The method according to claim 1, characterized in that The user satisfaction is optimized through the dynamic search result optimization mechanism to obtain and display optimized search results, including: Using the user satisfaction as a new ranking factor, and combining the new ranking factor with the traditional ranking factor to obtain a combined ranking factor; The search ranking algorithm is adjusted and optimized according to the combined ranking factors, so as to dynamically adjust the weight and ranking of each search result according to the user satisfaction score, and obtain and display optimized search results.

8. A user satisfaction analysis and processing device, characterized in that: The device comprises: A real-time acquisition unit, used to acquire user behavior data on the search result page in real time; A real-time data processing unit, used to perform real-time data processing on the user behavior data to obtain summary information; An evaluation unit, used to perform satisfaction evaluation through a pre-trained machine learning model and the summary information to obtain user satisfaction; An optimization unit, used to optimize the user satisfaction through a dynamic search result optimization mechanism, obtain optimized search results and display them; A training adjustment unit is used to train and adjust the machine learning model based on the optimized search results to form a closed-loop feedback system to ensure continuous optimization of the search results.

9. The device according to claim 8, characterized in that The real-time acquisition unit is specifically used for: The user behavior is tracked in real time to obtain the user behavior data on the search result page; wherein the user behavior data at least includes the number of clicks, dwell time, scroll depth, return frequency and sequence pattern analysis.

10. The device according to claim 8, characterized in that The real-time data processing at least includes data cleaning, feature extraction, data standardization, and real-time aggregation and summarization. The real-time data processing unit is specifically used for: The user behavior data is subjected to the data cleaning, feature extraction, data standardization and machine learning analysis to obtain summary information.

Images