Methods, apparatus, devices and storage media for optimizing gas sensor arrays

Abstract

This application relates to a gas sensor array optimization method, comprising: determining the feature importance of multiple candidate gas sensors in a gas model prediction within an electronic nose system; selecting a first target sensor from the multiple candidate gas sensors whose feature importance satisfies a high importance condition to form a sensor set; for each first target sensor in the sensor set, determining the interaction strength between the first target sensor and multiple remaining sensors, and selecting a second target sensor from the multiple remaining sensors whose interaction strength satisfies a high interaction condition to form a sensor subset of the first target sensor; determining a target sensor array based on the sensor set and the sensor subset; the target sensor array includes at least one first target sensor from the sensor set and one second target sensor from the sensor subset of the included first target sensor. This method can improve the performance of the electronic nose system.

Description

Technical Field

[0001] This application relates to the field of electronic nose technology, and in particular to a method, apparatus, device, and storage medium for optimizing a gas sensor array. Background Technology

[0002] Electronic nose systems, as biomimetic olfactory systems, are widely used in food safety, environmental protection, and disease diagnosis. To more realistically simulate human olfaction, artificial olfactory systems, i.e., electronic nose systems, typically employ multi-gas sensor arrays. However, too many gas sensors increase hardware complexity and lead to information redundancy due to cross-sensitivity; therefore, optimizing the gas sensor array is crucial.

[0003] Traditional gas sensor array optimization methods, such as those based on dynamic feature importance, genetic algorithms, feature selection based on correlation analysis, and machine learning, fail to balance the number of sensors and the performance of the gas model, resulting in poor performance of the electronic nose system. Summary of the Invention

[0004] Therefore, it is necessary to provide a gas sensor array optimization method, apparatus, device, and storage medium that can improve the performance of electronic nose systems, addressing the aforementioned technical problems.

[0005] In a first aspect, this application provides a method for optimizing a gas sensor array, the method comprising:

[0006] Determine the feature importance of multiple candidate gas sensors in the electronic nose system for gas model prediction;

[0007] From the plurality of candidate gas sensors, the first target sensor whose feature importance satisfies the high importance condition is selected to form a sensor set;

[0008] For each first target sensor in the sensor set, the interaction strength between the first target sensor and a plurality of remaining sensors is determined, and a second target sensor whose interaction strength satisfies the high interaction condition is selected from the plurality of remaining sensors to form a sensor subset of the first target sensor; the remaining sensors are candidate gas sensors other than the first target sensor among the plurality of candidate gas sensors.

[0009] A target sensor array is determined based on the sensor set and the sensor subset; the target sensor array includes at least one first target sensor from the sensor set and one second target sensor from the sensor subset containing the first target sensor.

[0010] In one embodiment, determining the feature importance of multiple candidate gas sensors in the electronic nose system for gas model prediction includes:

[0011] Identify all possible sensor arrays in the electronic nose system; each possible sensor array contains at least two candidate gas sensors from a plurality of candidate gas sensors in the electronic nose system.

[0012] For each candidate gas sensor in each of the possible sensor arrays, determine the importance of the candidate gas sensor in the gas model prediction, and obtain the initial feature importance of the candidate gas sensor in the possible sensor arrays;

[0013] For each of the plurality of candidate gas sensors, the initial feature importance of the candidate gas sensor in all possible sensor arrays is averaged to obtain the feature importance of the candidate gas sensor in gas model prediction.

[0014] In one embodiment, selecting a first target sensor from the plurality of candidate gas sensors whose feature importance satisfies the high importance condition to form a sensor set includes:

[0015] The candidate gas sensors are sorted in descending order of the importance of the features;

[0016] From the sorted candidate gas sensors, the first preset number of candidate gas sensors ranked first are selected as the first target sensors that meet the high importance condition, so as to form a sensor set.

[0017] In one embodiment, determining the interaction strength between each first target sensor and a plurality of remaining sensors for each first target sensor in the sensor set includes:

[0018] For each first target sensor in the sensor set, and for each remaining sensor in the plurality of remaining sensors, determine the joint probability between the first target sensor and the remaining sensors, and determine the first edge probability of the first target sensor and the second edge probability of the remaining sensors;

[0019] The interaction strength between the first target sensor and the remaining sensors is determined based on the joint probability, the first marginal probability, and the second marginal probability.

[0020] In one embodiment, selecting a second target sensor from the plurality of remaining sensors whose interaction intensity satisfies the high interaction condition to constitute a sensor subset of the first target sensor includes:

[0021] The remaining sensors are sorted in descending order of interaction intensity;

[0022] From the sorted plurality of remaining sensors, the second preset number of remaining sensors that are ranked first are selected as the second target sensors that meet the high interactivity condition, so as to form a sensor subset of the first target sensor.

[0023] In one embodiment, determining the target sensor array based on the parent set of sensors and the subset of sensors includes:

[0024] Based on the sensor set and the sensor subset, multiple candidate sensor arrays are sequentially determined using an incremental selection strategy. Each candidate sensor array contains the same number of candidate gas sensors in each subsequent determination. The number of candidate gas sensors in each subsequent determination is one more than the number of candidate gas sensors in the previous determination. Each candidate sensor array contains at least one first target sensor from the sensor set and one second target sensor from the sensor subset containing the first target sensor.

[0025] For each selection, the optimal sensor array that optimizes the model performance of the gas model is selected from the multiple candidate sensor arrays determined each time.

[0026] If the model performance increment between the optimal sensor array selected each time and the optimal sensor array selected last time is greater than or equal to a preset increment value, then the optimal sensor array that optimizes the model performance of the gas model will be selected from the multiple candidate sensor arrays determined in the next round. Otherwise, the selection will stop and the optimal sensor array selected last time will be used as the target sensor array.

[0027] Secondly, this application provides a gas sensor array optimization device, the device comprising:

[0028] The determination module is used to determine the feature importance of multiple candidate gas sensors in the electronic nose system in gas model prediction;

[0029] The selection module is used to select the first target sensor whose feature importance satisfies the high importance condition from the plurality of candidate gas sensors, so as to form a sensor set;

[0030] The determining module is further configured to determine the interaction strength between the first target sensor and a plurality of remaining sensors for each first target sensor in the sensor set; the selecting module is further configured to select a second target sensor from the plurality of remaining sensors whose interaction strength satisfies the high interaction condition, so as to form a sensor subset of the first target sensor; the remaining sensors are candidate gas sensors other than the first target sensor among the plurality of candidate gas sensors;

[0031] The determining module is further configured to determine a target sensor array based on the sensor set and the sensor subset; the target sensor array includes at least one first target sensor from the sensor set and one second target sensor from the sensor subset containing the first target sensor.

[0032] Thirdly, this application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the various method embodiments of this application.

[0033] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the various method embodiments of this application.

[0034] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps in the various method embodiments of this application.

[0035] The aforementioned gas sensor array optimization method, apparatus, device, and storage medium determine the feature importance of multiple candidate gas sensors in gas model prediction within an electronic nose system. From these candidate gas sensors, a first target sensor whose feature importance satisfies a high importance condition is selected to form a sensor subset. For each first target sensor in the sensor subset, the interaction strength between the first target sensor and multiple remaining sensors is determined, and a second target sensor whose interaction strength satisfies a high interaction condition is selected from the remaining sensors to form a sensor subset of the first target sensor. The remaining sensors are candidate gas sensors other than the first target sensor among the multiple candidate gas sensors. Based on the sensor subset and sensor subset, a target sensor array is determined. The target sensor array contains at least one first target sensor from the sensor subset and one second target sensor from the sensor subset containing the first target sensor. Compared to traditional gas sensor array optimization methods, this application first filters out core gas sensors from multiple gas sensors based on their feature importance in gas model prediction to retain key information. Then, by combining the interaction strength between the core gas sensor and other remaining sensors—a measure of the interdependence between two sensors—gas sensors with a high degree of dependence on the core gas sensor are selected from the remaining gas sensors. Furthermore, based on the selected core gas sensor and the gas sensors with a high degree of dependence on the core gas sensor, the final optimized target sensor array is determined. The optimized sensor array can ensure that the gas model performance remains at a good level while using fewer gas sensors, effectively balancing the number of sensors and the gas model performance, thereby improving the performance of the electronic nose system. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating a gas sensor array optimization method in one embodiment;

[0037] Figure 2 This is a flowchart illustrating the gas sensor array optimization method in another embodiment;

[0038] Figure 3 This is a schematic diagram showing a performance comparison between the optimized target sensor array of this application and a conventional array composed of all sensors in one embodiment;

[0039] Figure 4 This is a schematic diagram comparing the performance of the sensor array optimization method of this application with a traditional sensor array optimization method in one embodiment;

[0040] Figure 5This is a schematic diagram showing a performance comparison between the optimized target sensor array of this application and a conventional array composed of all sensors in another embodiment;

[0041] Figure 6 This is a structural block diagram of a gas sensor array optimization device in one embodiment;

[0042] Figure 7 This is an internal structural diagram of a computer device in one embodiment;

[0043] Figure 8 This is a diagram of the internal structure of a computer device in another embodiment. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0045] In one embodiment, such as Figure 1 As shown, a gas sensor array optimization method is provided. This method can be applied to computer devices, which can be terminals or servers. That is, the method can be executed independently by the terminal or the server, or it can be implemented through interaction between the terminal and the server. This embodiment illustrates the application of this method to a computer device as an example, including the following steps:

[0046] Step 102: Determine the feature importance of multiple candidate gas sensors in the electronic nose system in gas model prediction.

[0047] The electronic nose system consists of a gas sensor array and a gas model. The gas sensor array comprises multiple gas sensors. The gas sensor array is the hardware front-end of the electronic nose system, responsible for signal generation, while the gas model is the software back-end, responsible for signal parsing. Although the gas model itself does not contain the gas sensor array, its input depends entirely on the output of the gas sensor array. The feature importance of candidate gas sensors in gas model prediction refers to the degree to which a candidate gas sensor contributes to the gas model prediction result.

[0048] In one embodiment, the gas model can be at least one of a gas prediction model and a gas classification model. The gas prediction model is used to predict the input gas. The gas classification model is used to classify the input gas.

[0049] In one embodiment, determining the feature importance of multiple candidate gas sensors in an electronic nose system for gas model prediction includes: determining a candidate gas sensor array consisting of all candidate gas sensors in the electronic nose system; determining the importance of each candidate gas sensor in the candidate gas sensor array for gas model prediction, and directly using the determined importance as the feature importance of the candidate gas sensor in gas model prediction.

[0050] Step 104: Select the first target sensor from multiple candidate gas sensors whose feature importance satisfies the high importance condition to form a sensor set.

[0051] The first target sensor is at least one candidate gas sensor selected from multiple candidate gas sensors whose feature importance satisfies the high importance condition. The sensor set contains all the selected first target sensors.

[0052] In one embodiment, a computer device may select a first target sensor from a plurality of candidate gas sensors, whose feature importance is greater than a preset importance, to form a sensor set.

[0053] Step 106: For each first target sensor in the sensor set, determine the interaction strength between the first target sensor and multiple remaining sensors, and select a second target sensor from the multiple remaining sensors whose interaction strength satisfies the high interaction condition to form a sensor subset of the first target sensor; the remaining sensors are candidate gas sensors other than the first target sensor among multiple candidate gas sensors.

[0054] The interaction strength between the first target sensor and the remaining sensors measures the degree of interdependence between them. The second target sensor is at least one remaining sensor selected from a pool of remaining sensors whose interaction strength satisfies a high interaction condition. The sensor subset includes all selected second target sensors.

[0055] In one embodiment, for each first target sensor in the sensor set, the computer device can determine the interaction strength between the first target sensor and a plurality of remaining sensors, and select a second target sensor from the plurality of remaining sensors whose interaction strength is greater than a preset interaction strength to form a sensor subset of the first target sensor.

[0056] Step 108: Determine the target sensor array based on the sensor set and the sensor subset; the target sensor array contains at least one first target sensor from the sensor set and one second target sensor from the sensor subset containing the first target sensor.

[0057] In one embodiment, determining the target sensor array based on a sensor set and a sensor subset includes: determining a plurality of candidate sensor arrays based on the sensor set and the sensor subset; each candidate sensor array includes at least one first target sensor from the sensor set and one second target sensor from the sensor subset containing the first target sensor; and selecting the candidate sensor array that optimizes the model performance of the gas model from the plurality of candidate sensor arrays as the target sensor array.

[0058] In one embodiment, different evaluation metrics can be used to measure the performance of gas models for different task types. For example, for gas models performing classification tasks, gas classification accuracy can be used as the evaluation metric. For gas models performing regression tasks, mean squared error can be used as the evaluation metric.

[0059] In the aforementioned gas sensor array optimization method, the feature importance of multiple candidate gas sensors in the gas model prediction of the electronic nose system is determined. From the multiple candidate gas sensors, a first target sensor whose feature importance satisfies the high importance condition is selected to form a sensor set. For each first target sensor in the sensor set, the interaction strength between the first target sensor and multiple remaining sensors is determined, and from the multiple remaining sensors, a second target sensor whose interaction strength satisfies the high interaction condition is selected to form a sensor subset of the first target sensor. The remaining sensors are candidate gas sensors other than the first target sensor among the multiple candidate gas sensors. Based on the sensor set and sensor subset, a target sensor array is determined. The target sensor array contains at least one first target sensor from the sensor set and one second target sensor from the sensor subset of the included first target sensor. Compared to traditional gas sensor array optimization methods, this application first filters out core gas sensors from multiple gas sensors based on the feature importance of gas sensors in gas model prediction to retain key information. Then, combining the interaction strength between the core gas sensor and other remaining sensors as a measure of the interdependence between the two sensors, gas sensors with a high degree of dependence on the core gas sensor are filtered out from the remaining gas sensors. Furthermore, based on the selected core gas sensors and gas sensors that are highly dependent on the core gas sensors, the final optimized target sensor array is determined. The final optimized sensor array can ensure that the gas model performance remains at a good level with fewer gas sensors, thus balancing the number of sensors and the gas model performance well, thereby improving the performance of the electronic nose system.

[0060] In one embodiment, determining the feature importance of multiple candidate gas sensors in a gas model prediction within an electronic nose system includes: identifying all possible sensor arrays in the electronic nose system; each possible sensor array containing at least two candidate gas sensors from among the multiple candidate gas sensors in the electronic nose system; for each candidate gas sensor in each possible sensor array, determining the importance of the candidate gas sensor in the gas model prediction to obtain an initial feature importance of the candidate gas sensor in the possible sensor array; and for each candidate gas sensor among the multiple candidate gas sensors, averaging the initial feature importance of the candidate gas sensor across all possible sensor arrays to obtain the feature importance of the candidate gas sensor in the gas model prediction.

[0061] Among them, the possible sensor array is a sensor array that is a possible combination of all the candidate gas sensors in the electronic nose system.

[0062] In one embodiment, the initial feature importance (i.e. SHAP value) of each candidate gas sensor in each possible sensor array can be calculated using the following formula:

[0063] ∅ x = ∑ S⊆N(i) S ! N - S -1 ! N ! [f S⋃ i - f S ] ;

[0064] Where N is the set of sensors corresponding to the possible sensor array, S is the subset of sensors after removing candidate gas sensor i, f(S) represents the output of the gas model using the possible sensor array N for prediction, and f(S) represents the output of the gas model using only the sensor subset S for prediction. This represents the marginal contribution of gas sensor i.

[0065] In one embodiment, the feature importance of each candidate gas sensor in the gas model prediction can be calculated using the following formula:

[0066] ;

[0067] Where M is the number of samples, i.e., the number of possible sensor arrays. Let represent the initial feature importance of candidate gas sensor i in the possible sensor array j.

[0068] In the above embodiments, the influence of one gas sensor may change due to the presence of other gas sensors. Therefore, by averaging the importance of each gas sensor across all possible combinations of gas sensors, the feature importance of each gas sensor in gas model prediction can be obtained. This quantifies the average influence of each gas sensor on gas model prediction and improves the accuracy of the feature importance of each gas sensor in gas model prediction.

[0069] In one embodiment, selecting a first target sensor whose feature importance satisfies the high importance condition from a plurality of candidate gas sensors to form a sensor set includes: sorting the plurality of candidate gas sensors in descending order of feature importance; and selecting a first preset number of candidate gas sensors that are ranked first from the sorted plurality of candidate gas sensors as the first target sensor that satisfies the high importance condition to form a sensor set.

[0070] In the above embodiments, by selecting a subset of candidate gas sensors with the highest feature importance as the first target sensor that meets the high importance condition, and forming a sensor set, the core gas sensors can be screened more accurately to retain key information, thereby further improving the performance of the electronic nose system.

[0071] In one embodiment, for each first target sensor in the sensor set, determining the interaction strength between the first target sensor and a plurality of remaining sensors includes: for each first target sensor in the sensor set and for each remaining sensor in the plurality of remaining sensors, determining the joint probability between the first target sensor and the remaining sensors, and determining the first edge probability of the first target sensor and the second edge probability of the remaining sensors; and determining the interaction strength between the first target sensor and the remaining sensors based on the joint probability, the first edge probability and the second edge probability.

[0072] In one embodiment, the interaction strength between the first target sensor and the remaining sensors can be calculated using the following formula:

[0073] ;

[0074] Wherein, P(x,y) is the joint probability of simultaneously taking the value of the first target sensor x and the remaining sensor y, P(x) is the first marginal probability of taking the value of the first target sensor x, and P(y) is the second marginal probability of taking the value of the remaining sensor y.

[0075] In the above embodiments, the interaction strength between the first target sensor and the remaining sensors is determined by the joint probability between the first target sensor and the remaining sensors, the first edge probability of the first target sensor, and the second edge probability of the remaining sensors, which can improve the accuracy of the interaction strength.

[0076] In one embodiment, selecting a second target sensor from a plurality of remaining sensors that satisfies the high interaction intensity condition to form a sensor subset of the first target sensor includes: sorting the plurality of remaining sensors in descending order of interaction intensity; and selecting a second preset number of remaining sensors that are ranked first from the sorted plurality of remaining sensors as the second target sensors that satisfy the high interaction intensity condition to form a sensor subset of the first target sensor.

[0077] In the above embodiments, by using a subset of the remaining sensors with the highest interaction intensity as the second target sensors that meet the high importance condition, a sensor subset can be formed. This allows for more accurate screening of gas sensors that are highly dependent on the core gas sensor, thereby further improving the performance of the electronic nose system.

[0078] In one embodiment, determining the target sensor array based on a sensor set and a sensor subset includes: determining multiple candidate sensor arrays sequentially based on an incremental selection strategy according to the sensor set and the sensor subset; each candidate sensor array in each determined candidate sensor array contains the same number of candidate gas sensors; the number of candidate gas sensors in each determined candidate sensor array is one more than the number of candidate gas sensors in the previously determined candidate sensor array; each candidate sensor array contains at least one first target sensor from the sensor set and one second target sensor from the sensor subset containing the first target sensor; for each selection, selecting the optimal sensor array that optimizes the model performance of the gas model from the multiple candidate sensor arrays determined each time; if the model performance increment between the optimal sensor array selected each time and the optimal sensor array selected in the previous time is greater than or equal to a preset increment value, then continuing to select the optimal sensor array that optimizes the model performance of the gas model from the multiple candidate sensor arrays determined in the next time; otherwise, stopping the selection and using the optimal sensor array selected in the previous time as the target sensor array.

[0079] In the above embodiments, by incrementally selecting gas sensors and evaluating their contribution to model prediction, a gas sensor array that can better balance the number of sensors and the model performance of the gas model can be selected, thereby further improving the performance of the electronic nose system.

[0080] In one embodiment, such as Figure 2 As shown, the implementation steps of the gas sensor array optimization method of this application are as follows:

[0081] Step 1: Identify multiple candidate gas sensors S1-S8 in the electronic nose system.

[0082] Step 2: Determine the feature importance of multiple candidate gas sensors S1-S8 in the gas model prediction in the electronic nose system; sort the multiple candidate gas sensors in descending order of feature importance; select the first preset number (e.g., the first 2) of the candidate gas sensors ranked first from the sorted multiple candidate gas sensors as the first target sensors (S4 and S8) that meet the high importance condition, so as to form the sensor set.

[0083] Step 3: For each first target sensor in the sensor set, determine the interaction strength between the first target sensor and the multiple remaining sensors, and sort the multiple remaining sensors in descending order of interaction strength. From the sorted multiple remaining sensors, select the first few (e.g., the first 3) remaining sensors as the second target sensors that meet the high interaction condition, thus forming a sensor subset of the first target sensor. For example, the sensor subset of the first target sensor S4 in the sensor set includes S8, S1, and S5. The sensor subset of the first target sensor S8 in the sensor set includes S4, S5, and S1.

[0084] Steps 4 and 5: Based on the sensor set {S4, S8} and sensor subsets {S8, S1, S5} and {S4, S5, S1}, multiple candidate sensor arrays are sequentially determined using an incremental selection strategy. Each candidate sensor array contains the same number of candidate gas sensors. The number of candidate gas sensors in each determined candidate sensor array is one more than the number in the previous determined candidate sensor array. Each candidate sensor array contains at least one first target sensor from the sensor set and one second target sensor from the sensor subset containing the first target sensor. For each selection, the optimal sensor array that optimizes the gas model's performance is chosen from the multiple candidate sensor arrays determined each time. If the model performance increment between the optimal sensor array selected each time and the optimal sensor array selected in the previous time is greater than or equal to a preset increment value, then the next optimal sensor array that optimizes the gas model's performance is selected from the multiple candidate sensor arrays determined in the next time; otherwise, the selection stops, and the optimal sensor array selected in the previous time is taken as the target sensor array. For example, in step 5, the model performance increment between the optimal three-sensor array {S4, S8, S1} and the optimal two-sensor array {S4, S5} is... If the value is greater than or equal to the preset increment value β, then continue to select the optimal sensor array that optimizes the model performance of the gas model from the multiple candidate sensor arrays determined in the next round. Otherwise, stop the selection and take the previously selected optimal sensor array, i.e., {S4, S5}, as the target sensor array.

[0085] In one embodiment, such as Figure 3 As shown, Figure 3 In the diagram, A and B correspond to two datasets used for gas classification tasks. The performance of the target sensor array selected by the gas sensor array optimization method in this application on 11 different gas models (i.e., RF, DT, XGBoost, LR, SVM, LDA, CNN, RNN, LSTM, TCN, and MLP) is compared with the model performance and number of sensors of the original sensor array (i.e., the array composed of all gas sensors in the electronic nose system). The results clearly show that the gas sensor array optimization method in this application can not only reduce the number of gas sensors but also optimize model performance. Furthermore, the number of gas sensors corresponding to the target sensor array selected by the gas sensor array optimization method in this application is significantly reduced, by nearly half compared to the original sensor array. Based on RF, DT, XGBoost, SVM, and LDA models, the gas sensor array optimization method in this application can achieve array optimization using no more than two gas sensors while maintaining minimal fluctuations in model performance. In some gas models, the overall classification performance of the gas model is significantly improved as the number of gas sensors decreases.

[0086] In one embodiment, such as Figure 4 As shown, Figure 4 Figure A illustrates the classification confusion matrix performance of the gas sensor array optimization method of this application on different machine learning methods. It can be seen that, under different models, the gas sensor array optimization method of this application (i.e., the AOFI method) can effectively classify different categories. For the above classification task, the gas sensor array optimization method of this application was compared with other array optimization methods, and the evaluation criteria included accuracy, precision, recall, and F1 score. Figure 4 B and C in the figure show the performance comparison of the AOFI method with other array optimization methods on two datasets, and the performance of the original sensor array is marked as the baseline comparison. Figure 4 D and E in the diagram compare the number of gas sensors in the sensor arrays selected by different array optimization methods. By combining the comparison of model performance and the number of gas sensors, it can be seen that the AOFI method can achieve array optimization with a smaller number of gas sensors while maintaining model performance, thus achieving a good balance between the number of gas sensors and model performance.

[0087] In one embodiment, such as Figure 5 As shown, Figure 5 Figure A shows the regression performance test of four gases based on four data subsets and is validated by different models. It compares the MAE (mean absolute error) of the gas sensor array optimization method (i.e., AOFI method) in this application with that of the original sensor array combination. Figure 5 Figure B shows the R² performance of the gas sensor arrays selected using the AOFI method and other array optimization methods for each dataset. It can be seen that the AOFI method achieves excellent array optimization results across different datasets. Experimental results demonstrate that the AOFI method performs exceptionally well on regression tasks across multiple models. Compared to the original sensor array, the AOFI method improves the model's MAE and R² performance while reducing the number of sensors. For example, on models such as XGBoost, RF, and LSTM, the gas sensor array optimized by the AOFI method can achieve or even surpass the prediction performance of the original sensor array.

[0088] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially, these steps are not necessarily executed in that order. Unless otherwise expressly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the above embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0089] In one embodiment, such as Figure 6 As shown, a gas sensor array optimization device 600 is provided, which specifically includes:

[0090] The determination module 602 is used to determine the feature importance of multiple candidate gas sensors in the gas model prediction in the electronic nose system;

[0091] The selection module 604 is used to select the first target sensor whose feature importance satisfies the high importance condition from multiple candidate gas sensors to form a sensor set;

[0092] The determining module 602 is further configured to determine the interaction strength between the first target sensor and the plurality of remaining sensors for each first target sensor in the sensor set; the selecting module 604 is further configured to select a second target sensor from the plurality of remaining sensors whose interaction strength satisfies the high interaction condition, so as to form a sensor subset of the first target sensor; the remaining sensors are candidate gas sensors other than the first target sensor among the plurality of candidate gas sensors.

[0093] The determining module 602 is further configured to determine a target sensor array based on the sensor parent set and the sensor subset; the target sensor array includes at least one first target sensor from the sensor parent set and one second target sensor from the sensor subset containing the first target sensor.

[0094] In one embodiment, the determining module 602 is further configured to determine all possible sensor arrays in the electronic nose system; each possible sensor array includes at least two candidate gas sensors from a plurality of candidate gas sensors in the electronic nose system; for each candidate gas sensor in each possible sensor array, the importance of the candidate gas sensor in the gas model prediction is determined to obtain the initial feature importance of the candidate gas sensor in the possible sensor array; for each candidate gas sensor in the plurality of candidate gas sensors, the initial feature importance of the candidate gas sensor in all possible sensor arrays is averaged to obtain the feature importance of the candidate gas sensor in the gas model prediction.

[0095] In one embodiment, the selection module 604 is further configured to sort multiple candidate gas sensors in descending order of feature importance; and select the first preset number of candidate gas sensors ranked first from the sorted multiple candidate gas sensors as the first target sensor that meets the high importance condition, so as to form a sensor set.

[0096] In one embodiment, the determining module 602 is further configured to determine, for each first target sensor in the sensor set and for each of the plurality of remaining sensors, the joint probability between the first target sensor and the remaining sensors, and to determine the first edge probability of the first target sensor and the second edge probability of the remaining sensors; and to determine the interaction strength between the first target sensor and the remaining sensors based on the joint probability, the first edge probability and the second edge probability.

[0097] In one embodiment, the selection module 604 is further configured to sort a plurality of remaining sensors in descending order of interaction intensity; and select a second preset number of remaining sensors that are ranked first from the sorted plurality of remaining sensors as second target sensors that meet the high interaction condition, so as to form a sensor subset of the first target sensor.

[0098] In one embodiment, the determining module 602 is further configured to sequentially determine multiple candidate sensor arrays based on an incremental selection strategy according to a sensor parent set and a sensor subset; each candidate sensor array determined in each iteration contains the same number of candidate gas sensors; the number of candidate gas sensors in each determined candidate sensor array is one more than the number of candidate gas sensors in the previous determined candidate sensor array; each candidate sensor array contains at least one first target sensor from the sensor parent set and one second target sensor from the sensor subset containing the first target sensor; for each selection, the optimal sensor array that optimizes the model performance of the gas model is selected from the multiple candidate sensor arrays determined in each iteration; if the model performance increment between the optimal sensor array selected in each iteration and the optimal sensor array selected in the previous iteration is greater than or equal to a preset increment value, then the optimal sensor array that optimizes the model performance of the gas model is selected from the multiple candidate sensor arrays determined in the next iteration; otherwise, the selection is stopped, and the optimal sensor array selected in the previous iteration is taken as the target sensor array.

[0099] The aforementioned gas sensor array optimization device determines the feature importance of multiple candidate gas sensors in gas model prediction within an electronic nose system. From these candidate sensors, a first target sensor whose feature importance satisfies a high importance condition is selected to form a sensor subset. For each first target sensor in the sensor subset, the interaction strength between the first target sensor and multiple remaining sensors is determined, and a second target sensor whose interaction strength satisfies a high interaction condition is selected from the remaining sensors to form a sensor subset of the first target sensor. The remaining sensors are candidate gas sensors other than the first target sensor. Based on the sensor subset and sensor subset, a target sensor array is determined. The target sensor array contains at least one first target sensor from the sensor subset and one second target sensor from the sensor subset containing the first target sensor. Compared to traditional gas sensor array optimization methods, this application first filters out core gas sensors from multiple gas sensors based on their feature importance in gas model prediction to retain key information. Then, combining the interaction strength between the core gas sensor and other remaining sensors as a measure of the interdependence between the two sensors, gas sensors with a high degree of dependence on the core gas sensor are filtered out from the remaining gas sensors. Furthermore, based on the selected core gas sensors and gas sensors that are highly dependent on the core gas sensors, the final optimized target sensor array is determined. The final optimized sensor array can ensure that the gas model performance remains at a good level with fewer gas sensors, thus balancing the number of sensors and the gas model performance well, thereby improving the performance of the electronic nose system.

[0100] Each module in the aforementioned gas sensor array optimization device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0101] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 7As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and databases. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a gas sensor array optimization method.

[0102] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 8 As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a gas sensor array optimization method. The display unit of the computer device is used to form a visually visible image. It can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0103] Those skilled in the art will understand that Figure 7 and Figure 8The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0104] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0105] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0106] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0107] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0108] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0109] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0110] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for optimizing a gas sensor array, characterized in that, The method includes: Determine the feature importance of multiple candidate gas sensors in the electronic nose system for gas model prediction; From the plurality of candidate gas sensors, the first target sensor whose feature importance satisfies the high importance condition is selected to form a sensor set; For each first target sensor in the sensor set, the interaction strength between the first target sensor and a plurality of remaining sensors is determined, and a second target sensor whose interaction strength satisfies the high interaction condition is selected from the plurality of remaining sensors to form a sensor subset of the first target sensor; the remaining sensors are candidate gas sensors other than the first target sensor among the plurality of candidate gas sensors. A target sensor array is determined based on the sensor set and the sensor subset; the target sensor array includes at least one first target sensor from the sensor set and one second target sensor from the sensor subset containing the first target sensor.

2. The method according to claim 1, characterized in that, The determination of the feature importance of multiple candidate gas sensors in the electronic nose system in gas model prediction includes: Identify all possible sensor arrays in the electronic nose system; each possible sensor array contains at least two candidate gas sensors from a plurality of candidate gas sensors in the electronic nose system. For each candidate gas sensor in each of the possible sensor arrays, determine the importance of the candidate gas sensor in the gas model prediction, and obtain the initial feature importance of the candidate gas sensor in the possible sensor arrays; For each of the plurality of candidate gas sensors, the initial feature importance of the candidate gas sensor in all possible sensor arrays is averaged to obtain the feature importance of the candidate gas sensor in gas model prediction.

3. The method according to claim 1, characterized in that, The step of selecting a first target sensor from the plurality of candidate gas sensors that satisfies the high importance condition to form a sensor set includes: The candidate gas sensors are sorted in descending order of the importance of the features; From the sorted candidate gas sensors, the first preset number of candidate gas sensors ranked first are selected as the first target sensors that meet the high importance condition, so as to form a sensor set.

4. The method according to claim 1, characterized in that, For each first target sensor in the parent set of sensors, determining the interaction strength between the first target sensor and the plurality of remaining sensors includes: For each first target sensor in the sensor set, and for each remaining sensor in the plurality of remaining sensors, determine the joint probability between the first target sensor and the remaining sensors, and determine the first edge probability of the first target sensor and the second edge probability of the remaining sensors; The interaction strength between the first target sensor and the remaining sensors is determined based on the joint probability, the first marginal probability, and the second marginal probability.

5. The method according to claim 1, characterized in that, The step of selecting a second target sensor from the plurality of remaining sensors, wherein the interaction intensity satisfies the high interaction condition, to constitute a sensor subset of the first target sensor, includes: The remaining sensors are sorted in descending order of interaction intensity; From the sorted plurality of remaining sensors, the second preset number of remaining sensors that are ranked first are selected as the second target sensors that meet the high interactivity condition, so as to form a sensor subset of the first target sensor.

6. The method according to any one of claims 1 to 5, characterized in that, The step of determining the target sensor array based on the sensor set and the sensor subset includes: Based on the sensor set and the sensor subset, multiple candidate sensor arrays are sequentially determined using an incremental selection strategy. Each candidate sensor array contains the same number of candidate gas sensors in each subsequent determination. The number of candidate gas sensors in each subsequent determination is one more than the number of candidate gas sensors in the previous determination. Each candidate sensor array contains at least one first target sensor from the sensor set and one second target sensor from the sensor subset containing the first target sensor. For each selection, the optimal sensor array that optimizes the model performance of the gas model is selected from the multiple candidate sensor arrays determined each time. If the model performance increment between the optimal sensor array selected each time and the optimal sensor array selected last time is greater than or equal to a preset increment value, then the optimal sensor array that optimizes the model performance of the gas model will be selected from the multiple candidate sensor arrays determined in the next round. Otherwise, the selection will stop and the optimal sensor array selected last time will be used as the target sensor array.

7. A gas sensor array optimization device, characterized in that, The device includes: The determination module is used to determine the feature importance of multiple candidate gas sensors in the electronic nose system in gas model prediction; The selection module is used to select the first target sensor whose feature importance satisfies the high importance condition from the plurality of candidate gas sensors, so as to form a sensor set; The determining module is further configured to determine the interaction strength between the first target sensor and a plurality of remaining sensors for each first target sensor in the sensor set; the selecting module is further configured to select a second target sensor from the plurality of remaining sensors whose interaction strength satisfies the high interaction condition, so as to form a sensor subset of the first target sensor; the remaining sensors are candidate gas sensors other than the first target sensor among the plurality of candidate gas sensors; The determining module is further configured to determine a target sensor array based on the sensor set and the sensor subset; the target sensor array includes at least one first target sensor from the sensor set and one second target sensor from the sensor subset containing the first target sensor.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images