Predictive maintenance method and system for network equipment

Abstract

The invention discloses a predictive maintenance method and system for network equipment, and relates to the technical field of cloud computing networks. Comprising the steps of 1, collecting data of network equipment in real time, 2, preprocessing the data of the network equipment, 3, analyzing the preprocessed data, and constructing a prediction model: constructing a machine learning model, predicting the data, and predicting and processing time sequence data by using a deep learning model in combination with a long short-term memory (LSTM) network and a convolutional neural network (CNN); 4, testing and optimizing the prediction model: testing the prediction model in a controlled environment, and optimizing the model according to feedback; and 5, deploying the prediction model in the production environment, and monitoring the performance and accuracy of the prediction model.

Description

Technical Field

[0001] The invention discloses a predictive maintenance method and system for network equipment, and relates to the technical field of cloud computing network. Background Art

[0002] With the rise of technologies such as cloud computing, the Internet of Things (IoT), big data, and artificial intelligence (AI), the role of network devices has changed from a simple data transmission node to the core of intelligent management and decision support. In this context, the maintenance and management of network devices has become particularly critical. Existing network maintenance mainly relies on regular inspections and post-failure repairs. This passive maintenance strategy may be feasible when the network scale is small and there are fewer devices. However, with the expansion of network scale and the diversification of network services, this strategy can no longer meet the needs of modern enterprises. The increase in the number of network devices, the explosive growth of network traffic, and the strict requirements of service level agreements (SLAs) all require that network maintenance strategies must be more proactive and intelligent. Summary of the invention

[0003] In view of the problems of the prior art, the present invention provides a predictive maintenance method and system for network equipment, which monitors network equipment in a cloud environment in real time, uses data analysis and machine learning technology to analyze key performance indicators such as network traffic, load, latency, throughput, etc., predicts possible failures and performance bottlenecks, and helps operation and maintenance managers to achieve highly automated network equipment operation and maintenance.

[0004] The specific scheme proposed by the present invention is:

[0005] The present invention provides a predictive maintenance method for network equipment, comprising:

[0006] Step 1: Collect data from network devices in real time:

[0007] Collect data from third-party data sources: Obtain basic status information directly from network devices through device management protocols.

[0008] Collect monitoring traffic data: Capture network traffic data through network monitoring tools.

[0009] Collect load data: Obtain performance indicators of network devices through other device management interfaces.

[0010] Monitor and collect error logs of network devices;

[0011] Step 2: Preprocess the data of network devices:

[0012] Step 21: Clean the data.

[0013] Step 22: Divide the data into time windows, where continuous data is divided into fixed time windows.

[0014] Step 23: Smooth the data.

[0015] Step 24: Normalize the data.

[0016] Step 25: Perform seasonal adjustment on the data. If the data has obvious seasonal changes, seasonal adjustment should be performed.

[0017] Step 26: Process missing values.

[0018] Step 27: Perform data integration. If different data from network devices need to be combined for analysis, align the time of the data and integrate the different data into a unified data set.

[0019] Step 3: Analyze the preprocessed data:

[0020] Step 31: Perform data aggregation and time series analysis: Aggregate the data according to the time window and use time series analysis methods to identify trends and seasonal changes in the data.

[0021] Step 32: Perform anomaly detection analysis: Use statistical methods and machine learning algorithms to identify abnormal data patterns as early signal data for potential failures or performance bottlenecks, set thresholds, and automatically mark data anomalies when the data exceeds or falls below the threshold.

[0022] Step 33: Perform correlation analysis: Analyze the correlation between data and use covariance or correlation coefficient to quantify the linear relationship between different data indicators.

[0023] Step 34: Construct feature engineering: Extract features from the data, construct derived features, and use the features as important benchmarks for predictive models to construct feature engineering.

[0024] Step 35: Perform data pattern recognition: Use clustering algorithms to identify patterns or groups in the data.

[0025] Step 36: Build a prediction model: Build a machine learning model to predict data. Use a deep learning model to combine the long short-term memory network (LSTM) and the convolutional neural network (CNN) to predict and process time series data.

[0026] Step 4: Test and optimize the prediction model: Test the prediction model in a controlled environment and optimize the model based on the feedback;

[0027] Step 5: Deploy the predictive model in a production environment and monitor the performance and accuracy of the predictive model.

[0028] Furthermore, in step 31 of the predictive maintenance method for network equipment, a time series analysis method is used to identify the trend and seasonal changes of the throughput data using an autoregressive integrated moving average ARIMA model.

[0029] Furthermore, in step 32 of the predictive maintenance method for network equipment, statistical methods and isolation forest or One-Class SVM in machine learning algorithms are used to identify abnormal data patterns.

[0030] Furthermore, in step 34 of the predictive maintenance method for network equipment, features are extracted from the throughput data, the features include maximum value, minimum value, average value, standard deviation, and rate of change, and derived features are constructed. The derived features include data change trends and periodic fluctuations, and feature engineering is constructed using the features as important benchmarks for the prediction model.

[0031] Furthermore, in step 35 of the predictive maintenance method for network equipment, a clustering algorithm K-means or DBSCAN is used to identify patterns or groups in the throughput data.

[0032] Furthermore, in step 36 of the predictive maintenance method for network equipment, a machine learning model is constructed to predict data, wherein the machine learning model includes a random forest algorithm model, a gradient boosting machine GBM algorithm model, and a support vector machine SVM model.

[0033] Furthermore, in step 1 of the method for predictive maintenance of network equipment, when collecting monitoring flow data: using a network monitoring tool sFlow to capture network flow data,

[0034] When collecting load data, the performance indicators of network devices are obtained, including CPU utilization, memory usage, disk I / O,

[0035] When monitoring and collecting error logs of network devices, error logs include packet loss rate, error code, and connection failure information.

[0036] The present invention also provides a predictive maintenance system for network equipment, including a data acquisition module, a preprocessing module, a data analysis module, and a deployment module.

[0037] The data acquisition module collects data from network devices in real time:

[0038] Collect data from third-party data sources: Obtain basic status information directly from network devices through device management protocols.

[0039] Collect monitoring traffic data: Capture network traffic data through network monitoring tools.

[0040] Collect load data: Obtain performance indicators of network devices through other device management interfaces.

[0041] Monitor and collect error logs of network devices;

[0042] The preprocessing module preprocesses the data of network devices:

[0043] Step 21: Clean the data.

[0044] Step 22: Divide the data into time windows, where continuous data is divided into fixed time windows.

[0045] Step 23: Smooth the data.

[0046] Step 24: Normalize the data.

[0047] Step 25: Perform seasonal adjustment on the data. If the data has obvious seasonal changes, seasonal adjustment should be performed.

[0048] Step 26: Process missing values.

[0049] Step 27: Perform data integration. If different data from network devices need to be combined for analysis, align the time of the data and integrate the different data into a unified data set.

[0050] The data analysis module analyzes the preprocessed data:

[0051] Step 31: Perform data aggregation and time series analysis: Aggregate the data according to the time window and use time series analysis methods to identify trends and seasonal changes in the data.

[0052] Step 32: Perform anomaly detection analysis: Use statistical methods and machine learning algorithms to identify abnormal data patterns as early signal data for potential failures or performance bottlenecks, set thresholds, and automatically mark data anomalies when the data exceeds or falls below the threshold.

[0053] Step 33: Perform correlation analysis: Analyze the correlation between data and use covariance or correlation coefficient to quantify the linear relationship between different data indicators.

[0054] Step 34: Construct feature engineering: Extract features from the data, construct derived features, and use the features as important benchmarks for predictive models to construct feature engineering.

[0055] Step 35: Perform data pattern recognition: Use clustering algorithms to identify patterns or groups in the data.

[0056] Step 36: Build a prediction model: Build a machine learning model to predict data. Use a deep learning model to combine the long short-term memory network (LSTM) and the convolutional neural network (CNN) to predict and process time series data.

[0057] Data analysis module tests and optimizes prediction models: Tests prediction models in a controlled environment and optimizes models based on feedback;

[0058] The deployment module deploys the predictive model in a production environment and monitors the performance and accuracy of the predictive model.

[0059] The benefits of the present invention are:

[0060] Improve network stability: Through real-time monitoring and analysis of network device throughput, potential performance issues and failures can be discovered in a timely manner. By predicting the time when a failure will occur, the system can take measures in advance, such as adjusting network configuration or replacing hardware, to avoid failures and significantly improve network stability and reliability.

[0061] Reduced maintenance costs: Predictive maintenance allows maintenance to be performed before problems occur, thereby reducing the need for emergency repairs, reducing maintenance costs, and helping to optimize inventory management as spare parts requirements can be more accurately predicted.

[0062] Optimize resource allocation: By analyzing the data, network traffic patterns in different time periods can be predicted to help network administrators allocate network resources more effectively. For example, bandwidth can be increased or routing policies can be adjusted before high traffic periods to ensure that network performance is not affected. This dynamic resource management can improve the overall efficiency and responsiveness of the network.

[0063] Improve data-driven decision-making capabilities: By collecting and analyzing large amounts of network performance data, the decision-making process can be made more data-driven. This data-based approach can reduce human bias and improve the accuracy and efficiency of decision-making. At the same time, customized reports and dashboards can be provided to help managers at different levels quickly obtain key information. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a schematic diagram of the interaction of system modules of the present invention. DETAILED DESCRIPTION

[0065] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0066] Example 1

[0067] The present invention provides a predictive maintenance method for network equipment, comprising:

[0068] Step 1: Collect data from network devices in real time:

[0069] Collect data through third-party data sources: Obtain basic status information directly from network devices through device management protocols such as SSH and Telnet. Basic status information includes firmware version, configuration parameters, interface status, etc.

[0070] Collect and monitor traffic data: Capture network traffic data through the network monitoring tool sFlow, which provides detailed information on data packets transmitted in the network, including source address, destination address, protocol type, transmission rate, etc.

[0071] Collect load data: Obtain performance indicators of network devices through other device management interfaces, such as CPU utilization, memory usage, disk I / O, etc. to evaluate device load.

[0072] Monitor and collect error logs of network devices, including packet loss rate, error code, connection failure, etc. Data can be collected through the monitoring system integrated with Zabbix to collect device log system data.

[0073] Step 2: Preprocess the data of network devices, taking throughput data as an example:

[0074] Step 21: Clean the data to identify and process outliers in the throughput data. Outliers may be caused by transient network congestion, equipment failure, or other abnormal conditions. Threshold detection or statistical-based methods (such as standard deviation, box plots) can be used to identify these outliers and replace or delete them.

[0075] Step 22: Divide the data into time windows, where the continuous throughput data is divided into fixed time windows, such as every 5 minutes or every 15 minutes, to facilitate the analysis of network performance within a specific time period. This helps to discover periodic changes or trends.

[0076] Step 23: Smooth the data. Smooth the throughput data to reduce the impact of random fluctuations. Commonly used smoothing techniques include moving average, exponential smoothing, etc.

[0077] Step 24: Normalize the data. If the throughput data is analyzed together with other network performance indicators, such as latency and packet loss rate, they need to be normalized to the same level to facilitate model training and comparison. Common normalization methods include minimum-maximum normalization or Z-score normalization.

[0078] Step 25: Perform seasonal adjustment on the data. If the data has obvious seasonal changes, seasonal adjustment should be performed.

[0079] Step 26: Perform missing value processing. If there are missing values ​​in the throughput data, it is necessary to decide whether to fill the missing values ​​or delete the corresponding data points based on the context and data characteristics.

[0080] Step 27: Perform data integration. If different data from network devices need to be combined for analysis, align the time of the data and integrate the different data into a unified data set. If the throughput data needs to be combined with other network performance indicators for analysis, ensure that the data is time-aligned and integrate them into a unified data set.

[0081] Step 3: Analyze the pre-processed throughput data:

[0082] Step 31: Perform data aggregation and time series analysis: Extract useful features from the throughput data, such as peak throughput, average throughput, throughput change rate, etc. These features can provide deeper insights into network performance. Aggregate the throughput data by time window, such as hourly and daily, and use time series analysis methods to identify trends and seasonal changes in the data through the autoregressive integrated moving average (ARIMA) model.

[0083] Step 32: Perform anomaly detection analysis: Use statistical methods and machine learning algorithms such as Isolation Forest and One-Class SVM to identify abnormal throughput data patterns as early signal data of potential failures or performance bottlenecks, set thresholds, and automatically mark data anomalies when the data exceeds or falls below the threshold.

[0084] Step 33: Perform correlation analysis: Analyze the correlation between data and use covariance or correlation coefficient to quantify the linear relationship between different data indicators.

[0085] Step 34: Construct feature engineering: Extract features from the throughput data, such as maximum value, minimum value, average value, standard deviation, rate of change, etc., construct derived features, such as the throughput change trend, periodic fluctuations, etc., and use the features as an important benchmark for the prediction model to construct feature engineering.

[0086] Step 35: Perform data pattern recognition: Use clustering algorithms such as K-means and DBSCAN to identify patterns or groups in the data.

[0087] Step 36: Build a prediction model: Build a machine learning model, including random forest, gradient boosting machine (GBM), support vector machine (SVM), etc., to predict data, and use a deep learning model, combining the long short-term memory network LSTM and the convolutional neural network CNN, to predict and process time series data.

[0088] Step 4: Test and optimize the prediction model: Test the prediction model in a controlled environment and optimize the model based on the feedback;

[0089] Step 5: Deploy the prediction model in the production environment and monitor the performance and accuracy of the prediction model. Make data decisions based on the prediction data of the prediction model. The implementation process can be referred to as follows:

[0090] Fault prediction: Based on historical throughput data and prediction models, predict the time points of possible future faults. Set up an early warning system to notify network administrators in a timely manner when the prediction results show potential faults.

[0091] Performance bottleneck identification: Analyze throughput data to identify the conditions and patterns where performance bottlenecks occur. Provide performance optimization suggestions, such as increasing bandwidth and optimizing routing strategies.

[0092] Resource scheduling: Dynamically adjust network resource allocation based on throughput prediction results to meet traffic demands in different time periods. Implement load balancing strategies to avoid overload and performance degradation.

[0093] Maintenance plan formulation: Combine failure prediction and performance bottleneck analysis to formulate preventive maintenance plans. Determine the maintenance time window to minimize the impact on business.

[0094] Simulation testing: Test congestion management and failover strategies in a simulated environment.

[0095] You can also deploy and maintain the system: deploy congestion management and failover systems in production networks. Continuously monitor system performance, regularly maintain and update policies.

[0096] User interface and interaction: Design an intuitive user interface to display throughput data, prediction results, and maintenance recommendations. Provide interactive tools to allow users to adjust prediction models and maintenance strategies based on actual conditions.

[0097] Through the above data analysis and decision support methods, predictive maintenance of network equipment can help operation and maintenance personnel identify and solve performance problems of network equipment in advance, thereby improving the stability and reliability of the network.

[0098] Example 2

[0099] The present invention also provides a predictive maintenance system for network equipment, including a data acquisition module, a preprocessing module, a data analysis module, and a deployment module.

[0100] The data acquisition module collects data from network devices in real time:

[0101] Collect data from third-party data sources: Obtain basic status information directly from network devices through device management protocols.

[0102] Collect monitoring traffic data: Capture network traffic data through network monitoring tools.

[0103] Collect load data: Obtain performance indicators of network devices through other device management interfaces.

[0104] Monitor and collect error logs of network devices;

[0105] The preprocessing module preprocesses the data of network devices:

[0106] Step 21: Clean the data.

[0107] Step 22: Divide the data into time windows, where continuous data is divided into fixed time windows.

[0108] Step 23: Smooth the data.

[0109] Step 24: Normalize the data.

[0110] Step 25: Perform seasonal adjustment on the data. If the data has obvious seasonal changes, seasonal adjustment should be performed.

[0111] Step 26: Process missing values.

[0112] Step 27: Perform data integration. If different data from network devices need to be combined for analysis, align the time of the data and integrate the different data into a unified data set.

[0113] The data analysis module analyzes the preprocessed data:

[0114] Step 31: Perform data aggregation and time series analysis: Aggregate the data according to the time window and use time series analysis methods to identify trends and seasonal changes in the data.

[0115] Step 32: Perform anomaly detection analysis: Use statistical methods and machine learning algorithms to identify abnormal data patterns as early signal data for potential failures or performance bottlenecks, set thresholds, and automatically mark data anomalies when the data exceeds or falls below the threshold.

[0116] Step 33: Perform correlation analysis: Analyze the correlation between data and use covariance or correlation coefficient to quantify the linear relationship between different data indicators.

[0117] Step 34: Construct feature engineering: Extract features from the data, construct derived features, and use the features as important benchmarks for predictive models to construct feature engineering.

[0118] Step 35: Perform data pattern recognition: Use clustering algorithms to identify patterns or groups in the data.

[0119] Step 36: Build a prediction model: Build a machine learning model to predict data. Use a deep learning model to combine the long short-term memory network (LSTM) and the convolutional neural network (CNN) to predict and process time series data.

[0120] Data analysis module tests and optimizes prediction models: Tests prediction models in a controlled environment and optimizes models based on feedback;

[0121] The deployment module deploys the predictive model in a production environment and monitors the performance and accuracy of the predictive model.

[0122] The information interaction, execution process and other contents between the modules in the above-mentioned system are based on the same concept as the embodiment of the method of the present invention. For the specific contents, please refer to the description in the embodiment of the method of the present invention, and no further description is given here.

[0123] Likewise, the benefits of the system of the present invention are:

[0124] Improve network stability: Through real-time monitoring and analysis of network device throughput, potential performance issues and failures can be discovered in a timely manner. By predicting the time when a failure will occur, the system can take measures in advance, such as adjusting network configuration or replacing hardware, to avoid failures and significantly improve network stability and reliability.

[0125] Reduced maintenance costs: Predictive maintenance allows maintenance to be performed before problems occur, thereby reducing the need for emergency repairs, reducing maintenance costs, and helping to optimize inventory management as spare parts requirements can be more accurately predicted.

[0126] Optimize resource allocation: By analyzing the data, network traffic patterns in different time periods can be predicted to help network administrators allocate network resources more effectively. For example, bandwidth can be increased or routing policies can be adjusted before high traffic periods to ensure that network performance is not affected. This dynamic resource management can improve the overall efficiency and responsiveness of the network.

[0127] Improve data-driven decision-making capabilities: By collecting and analyzing large amounts of network performance data, the decision-making process can be made more data-driven. This data-based approach can reduce human bias and improve the accuracy and efficiency of decision-making. At the same time, customized reports and dashboards can be provided to help managers at different levels quickly obtain key information.

[0128] It should be noted that not all steps and modules in the above-mentioned processes and system structures are necessary, and some steps or modules can be ignored according to actual needs. The execution order of each step is not fixed and can be adjusted as needed. The system structure described in the above-mentioned embodiments can be a physical structure or a logical structure, that is, some modules may be implemented by the same physical entity, or some modules may be implemented by multiple physical entities, or some components in multiple independent devices may be implemented together.

[0129] The above-described embodiments are only preferred embodiments for fully illustrating the present invention, and the protection scope of the present invention is not limited thereto. Equivalent substitutions or changes made by those skilled in the art based on the present invention are within the protection scope of the present invention. The protection scope of the present invention shall be subject to the claims.

Claims

1. A predictive maintenance method for network equipment, characterized in that include: Step 1: Collect data from network devices in real time: Collect data from third-party data sources: Obtain basic status information directly from network devices through device management protocols. Collect monitoring traffic data: Capture network traffic data through network monitoring tools. Collect load data: Obtain performance indicators of network devices through other device management interfaces. Monitor and collect error logs of network devices; Step 2: Preprocess the data of network devices: Step 21: Clean the data. Step 22: Divide the data into time windows, where continuous data is divided into fixed time windows. Step 23: Smooth the data. Step 24: Normalize the data. Step 25: Perform seasonal adjustment on the data. If the data has obvious seasonal changes, seasonal adjustment should be performed. Step 26: Process missing values. Step 27: Perform data integration. If different data from network devices need to be combined for analysis, align the time of the data and integrate the different data into a unified data set. Step 3: Analyze the preprocessed data: Step 31: Perform data aggregation and time series analysis: Aggregate the data according to the time window and use time series analysis methods to identify trends and seasonal changes in the data. Step 32: Perform anomaly detection analysis: Use statistical methods and machine learning algorithms to identify abnormal data patterns as early signal data for potential failures or performance bottlenecks, set thresholds, and automatically mark data as abnormal when the data exceeds or falls below the threshold. Step 33: Perform correlation analysis: Analyze the correlation between data and use covariance or correlation coefficient to quantify the linear relationship between different data indicators. Step 34: Construct feature engineering: Extract features from the data, construct derived features, and use the features as important benchmarks for predictive models to construct feature engineering. Step 35: Perform data pattern recognition: Use clustering algorithms to identify patterns or groups in the data. Step 36: Build a prediction model: Build a machine learning model to predict data. Use a deep learning model to combine the long short-term memory network (LSTM) and the convolutional neural network (CNN) to predict and process time series data. Step 4: Test and optimize the prediction model: Test the prediction model in a controlled environment and optimize the model based on the feedback; Step 5: Deploy the predictive model in a production environment and monitor the performance and accuracy of the predictive model.

2. A predictive maintenance method for network equipment according to claim 1, characterized in that In step 31, a time series analysis method is used to identify the trend and seasonal changes of the throughput data using an autoregressive integrated moving average (ARIMA) model.

3. A predictive maintenance method for network equipment according to claim 1, characterized in that In step 32, statistical methods and isolation forest or One-Class SVM in machine learning algorithms are used to identify abnormal data patterns.

4. A predictive maintenance method for network equipment according to claim 1, characterized in that In step 34, features are extracted from the throughput data. The features include the maximum value, minimum value, average value, standard deviation, and rate of change. Derivative features are constructed. The derived features include data change trends and periodic fluctuations. The features are used as important benchmarks for the prediction model to construct feature engineering.

5. The predictive maintenance method for network equipment according to claim 1, characterized in that In step 35, a clustering algorithm such as K-means or DBSCAN is used to identify patterns or groups in the throughput data.

6. According to a predictive maintenance method for network equipment according to claim 1, it is characterized in that a machine learning model is constructed in step 36 to predict data, wherein the machine learning model includes a random forest algorithm model, a gradient boosting machine GBM algorithm model, and a support vector machine SVM model.

7. A predictive maintenance method for network equipment according to claim 1, characterized in that When collecting monitoring traffic data in step 1: Use the network monitoring tool sFlow to capture network traffic data. When collecting load data, the performance indicators of network devices are obtained, including CPU utilization, memory usage, disk I / O, When monitoring and collecting error logs of network devices, error logs include packet loss rate, error code, and connection failure information.

8. A predictive maintenance system for network equipment, characterized in that Including data acquisition module, preprocessing module, data analysis module, deployment module, The data acquisition module collects data from network devices in real time: Collect data from third-party data sources: Obtain basic status information directly from network devices through device management protocols. Collect monitoring traffic data: Capture network traffic data through network monitoring tools. Collect load data: Obtain performance indicators of network devices through other device management interfaces. Monitor and collect error logs of network devices; The preprocessing module preprocesses the data of network devices: Step 21: Clean the data. Step 22: Divide the data into time windows, where continuous data is divided into fixed time windows. Step 23: Smooth the data. Step 24: Normalize the data. Step 25: Perform seasonal adjustment on the data. If the data has obvious seasonal changes, seasonal adjustment should be performed. Step 26: Process missing values. Step 27: Perform data integration. If different data from network devices need to be combined for analysis, align the time of the data and integrate the different data into a unified data set. The data analysis module analyzes the preprocessed data: Step 31: Perform data aggregation and time series analysis: Aggregate the data according to the time window and use time series analysis methods to identify trends and seasonal changes in the data. Step 32: Perform anomaly detection analysis: Use statistical methods and machine learning algorithms to identify abnormal data patterns as early signal data for potential failures or performance bottlenecks, set thresholds, and automatically mark data as abnormal when the data exceeds or falls below the threshold. Step 33: Perform correlation analysis: Analyze the correlation between data and use covariance or correlation coefficient to quantify the linear relationship between different data indicators. Step 34: Construct feature engineering: Extract features from the data, construct derived features, and use the features as important benchmarks for predictive models to construct feature engineering. Step 35: Perform data pattern recognition: Use clustering algorithms to identify patterns or groups in the data. Step 36: Build a prediction model: Build a machine learning model to predict data. Use a deep learning model to combine the long short-term memory network (LSTM) and the convolutional neural network (CNN) to predict and process time series data. Data analysis module tests and optimizes prediction models: Tests prediction models in a controlled environment and optimizes models based on feedback; The deployment module deploys the predictive model in a production environment and monitors the performance and accuracy of the predictive model.

Images