AI-based wastewater biological toxicity prediction method and system

Abstract

The present application relates to the field of wastewater detection technology, and provides a wastewater biological toxicity prediction method and system based on AI, which includes obtaining molecular characteristic data information of the wastewater to be tested, and generating a fingerprint characteristic matrix corresponding to the molecular characteristic data information; determining the toxicity level of each characteristic compound in the fingerprint characteristic matrix, and clustering and classifying each characteristic compound based on the toxicity level of each characteristic compound to obtain multiple toxicity category sets; determining the information gain corresponding to each fingerprint position in the fingerprint characteristic matrix based on each toxicity category set; generating a key characteristic matrix based on the information gain corresponding to each fingerprint position; inputting the key characteristic matrix into a preset integrated biological toxicity prediction model to obtain the biological toxicity prediction result of the wastewater to be tested. This method realizes effective feature screening of molecular characteristic data information, which helps to improve the accuracy of wastewater biological toxicity prediction results.

Description

Technical Field

[0001] The present application relates to the field of wastewater detection technology, and in particular to an AI-based wastewater biological toxicity prediction method and system. Background Art

[0002] Wastewater biological toxicity refers to the toxic effects of toxic and harmful substances contained in wastewater on organisms. With the acceleration of urban industrialization and the growth of urban population, the discharge of industrial wastewater and domestic sewage has increased dramatically, which contains various complex organic matter, heavy metals, new pollutants, etc. Even after treatment, some trace toxic and harmful substances may be converted into intermediate products with stronger biological toxicity, resulting in the inability to effectively control the biological toxicity of wastewater. These toxic and harmful substances may pose a serious threat to aquatic ecosystems and human health. Therefore, the detection and evaluation of wastewater biological toxicity has important environmental significance and social value.

[0003] With the development of artificial intelligence (AI) and machine learning technology, the application of AI in the field of wastewater biological toxicity prediction has become a research hotspot. However, existing wastewater biological toxicity prediction methods usually directly use the molecular characteristic data information in the wastewater to be tested for toxicity prediction, lacking effective feature screening methods, resulting in data redundancy and reducing prediction accuracy. Summary of the invention

[0004] This application provides an AI-based wastewater biological toxicity prediction method and system to solve the problems raised by the above background technology.

[0005] In a first aspect, the present application provides an AI-based wastewater biological toxicity prediction method, comprising:

[0006] Acquire molecular characteristic data information of the wastewater to be tested, and generate a fingerprint characteristic matrix corresponding to the molecular characteristic data information based on a preset molecular fingerprint algorithm;

[0007] Determine the toxicity level of each characteristic compound in the fingerprint characteristic matrix based on the chemical substance toxicity judgment standard, and cluster and classify each characteristic compound based on the toxicity level of each characteristic compound to obtain multiple toxicity category sets; wherein one toxicity category set corresponds to one toxicity level;

[0008] Determine the information gain corresponding to each fingerprint bit in the fingerprint feature matrix based on each of the toxicity category sets;

[0009] Generate an initial key feature matrix based on the information gain corresponding to each of the fingerprint positions;

[0010] Performing dimensionality reduction processing on the initial key feature matrix to obtain a key feature matrix;

[0011] The key feature matrix is ​​input into a preset integrated biological toxicity prediction model to obtain a biological toxicity prediction result of the wastewater to be tested.

[0012] In a possible implementation, determining the information gain corresponding to each fingerprint bit in the fingerprint feature matrix based on each of the toxicity category sets includes:

[0013] Calculating the conditional probabilities corresponding to each of the toxicity category sets respectively, and obtaining the overall entropy based on the conditional probabilities corresponding to each of the toxicity category sets;

[0014] Determine the conditional entropy corresponding to each of the fingerprint positions respectively;

[0015] For each of the fingerprint bits, the conditional entropy corresponding to the fingerprint bit is subtracted from the overall entropy to obtain the information gain corresponding to the fingerprint bit.

[0016] In a possible implementation, performing dimensionality reduction processing on the initial key feature matrix to obtain the key feature matrix includes:

[0017] Performing binary combination on each initial key fingerprint position in the initial key feature matrix to obtain a plurality of binary combination pairs;

[0018] For each of the binary combination pairs, calculating the Pearson correlation coefficient between the two initial key fingerprint positions of the binary combination pair;

[0019] For each of the binary combination pairs, if the Pearson correlation coefficient corresponding to the binary combination pair is greater than a preset threshold, the column vector corresponding to the initial key fingerprint position with smaller information gain in the binary combination pair is deleted from the initial key feature matrix.

[0020] In a possible implementation, the training method of the integrated biological toxicity prediction model includes:

[0021] Obtaining a training data set; wherein the training data set includes a plurality of samples, each of which includes a feature matrix and a corresponding biological toxicity standard value;

[0022] Dividing the training data set into a training set, a validation set and a test set;

[0023] Based on the training set, each preset basic learner is trained respectively to obtain a sub-model corresponding to each basic learner;

[0024] For each of the sub-models, judging whether the model parameters of the sub-model need to be adjusted based on the validation set; if necessary, adjusting the model parameters of the sub-model based on a preset parameter adjustment method to obtain a target sub-model corresponding to the sub-model; if not necessary, determining the sub-model as the target sub-model;

[0025] Determine the weight coefficient of each target sub-model based on the test set;

[0026] The integrated biological toxicity prediction model is constructed based on the weight coefficients of each target sub-model.

[0027] In a possible implementation, the determining whether the model parameters of the sub-model need to be adjusted based on the validation set includes:

[0028] For each verification sample in the verification set, input the feature matrix corresponding to the verification sample into the sub-model to obtain the biological toxicity prediction value corresponding to the verification sample, and obtain the first absolute value of the difference between the biological toxicity standard value corresponding to the verification sample and the biological toxicity prediction value;

[0029] respectively comparing each of the first absolute values ​​with a first preset absolute value;

[0030] If any of the first absolute values ​​is greater than the first preset absolute value, it is determined that the model parameters of the sub-model need to be adjusted.

[0031] In a possible implementation, adjusting the model parameters of the sub-model based on a preset parameter adjustment method to obtain a target sub-model corresponding to the sub-model includes:

[0032] For each of the verification samples, when the first absolute value corresponding to the verification sample is greater than the first preset absolute value, a binary cross entropy loss function value is constructed based on the biological toxicity standard value and the biological toxicity prediction value corresponding to the verification sample, and the output error value is obtained by subtracting the biological toxicity prediction value from the biological toxicity standard value, and based on the binary cross entropy loss function value and the output error value, the model parameters of the sub-model are adjusted using the back propagation method to obtain the target sub-model corresponding to the sub-model.

[0033] In a possible implementation manner, determining the weight coefficient of each target sub-model based on the test set includes:

[0034] For each of the target sub-models, determining the prediction accuracy of the target sub-model based on the test set;

[0035] Adding the prediction accuracy of each target sub-model to obtain the sum of the prediction accuracy;

[0036] For each of the target sub-models, the ratio of the prediction accuracy rate corresponding to the target sub-model to the sum of the prediction accuracy rates is used as the weight coefficient of the target sub-model.

[0037] In a possible implementation manner, determining the prediction accuracy of the target sub-model based on the test set includes:

[0038] For each test sample in the test set, input the feature matrix corresponding to the test sample into the target sub-model to obtain the biological toxicity prediction value corresponding to the test sample, and obtain the second absolute value of the difference between the biological toxicity standard value corresponding to the test sample and the biological toxicity prediction value;

[0039] Compare each of the second absolute values ​​with a second preset absolute value respectively; wherein the second preset absolute value is smaller than the first preset absolute value;

[0040] If the second absolute value is not greater than the second preset absolute value, determining that the test sample corresponding to the second absolute value is an accurate test sample;

[0041] The number of accurate test samples is counted, and the ratio between the number of accurate test samples and the total number of samples in the test set is determined as the prediction accuracy of the target sub-model.

[0042] In a second aspect, the present application provides an AI-based wastewater biological toxicity prediction system, comprising:

[0043] An acquisition module is used to acquire molecular characteristic data information of the wastewater to be tested, and generate a fingerprint characteristic matrix corresponding to the molecular characteristic data information based on a preset molecular fingerprint algorithm;

[0044] A clustering classification module, used to determine the toxicity level of each characteristic compound in the fingerprint characteristic matrix based on the chemical substance toxicity judgment standard, and cluster and classify each characteristic compound based on the toxicity level of each characteristic compound to obtain multiple toxicity category sets; wherein one toxicity category set corresponds to one toxicity level;

[0045] A determination module, configured to determine the information gain corresponding to each fingerprint position in the fingerprint feature matrix based on each of the toxicity category sets;

[0046] A generating module, used for generating an initial key feature matrix based on the information gain corresponding to each of the fingerprint positions;

[0047] A dimension reduction processing module is used to perform dimension reduction processing on the initial key feature matrix to obtain a key feature matrix;

[0048] The input module is used to input the key feature matrix into a preset integrated biological toxicity prediction model to obtain the biological toxicity prediction result of the wastewater to be tested.

[0049] The present application provides an AI-based wastewater biological toxicity prediction method and system, which includes: obtaining molecular characteristic data information of the wastewater to be tested, and generating a fingerprint characteristic matrix corresponding to the molecular characteristic data information based on a preset molecular fingerprint algorithm; determining the toxicity level of each characteristic compound in the fingerprint characteristic matrix based on the chemical substance toxicity judgment standard, and clustering and classifying each characteristic compound based on the toxicity level of each characteristic compound to obtain a plurality of toxicity category sets; wherein, one toxicity category set corresponds to one toxicity level; determining the information gain corresponding to each fingerprint position in the fingerprint characteristic matrix based on each toxicity category set; generating an initial key feature matrix based on the information gain corresponding to each fingerprint position; performing dimensionality reduction processing on the initial key feature matrix to obtain a key feature matrix; inputting the key feature matrix into a preset integrated biological toxicity prediction model to obtain the biological toxicity prediction result of the wastewater to be tested. The method, on the one hand, determines the information gain corresponding to each fingerprint position in the fingerprint feature matrix based on each toxicity category set; generates an initial key feature matrix based on the information gain corresponding to each fingerprint position; performs dimensionality reduction processing on the initial key feature matrix to obtain a key feature matrix, thereby achieving effective feature screening of molecular feature data information, which helps to improve the accuracy of wastewater biological toxicity prediction results. On the other hand, by inputting the key feature matrix into a preset integrated biological toxicity prediction model, the biological toxicity prediction result of the wastewater to be tested is obtained, thereby further improving the accuracy of the wastewater biological toxicity prediction result. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0051] Figure 1 A schematic diagram of a wastewater biological toxicity prediction method based on AI provided in an embodiment of the present application;

[0052] Figure 2 A schematic block diagram of the structure of the wastewater biological toxicity prediction system based on AI provided in an embodiment of the present application;

[0053] Figure 3 A schematic block diagram of the structure of a terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0056] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0057] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0058] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0059] See also Figure 1 , Figure 1 A schematic diagram of a wastewater biological toxicity prediction method based on AI provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the AI-based wastewater biological toxicity prediction method provided in the embodiment of the present application includes steps S1 to S6.

[0060] Step S1, obtaining molecular characteristic data information of the wastewater to be tested, and generating a fingerprint characteristic matrix corresponding to the molecular characteristic data information based on a preset molecular fingerprint algorithm.

[0061] Specifically, the wastewater to be tested is detected by an ultra-high performance liquid chromatography-high resolution mass spectrometry system to obtain a total ion current chromatogram of the wastewater to be tested, and the total ion current chromatogram is analyzed to obtain the molecular characteristic data information of the wastewater to be tested, wherein the molecular fingerprint algorithm is a cyclic fingerprint algorithm ECFP4, which can flexibly capture various local structural features based on the atomic environment, and has good distinguishing ability for molecules with structural diversity. The rows of the fingerprint feature matrix represent characteristic compounds, and the columns represent fingerprint bits, and each fingerprint bit corresponds to a specific structural fragment, substructure or feature in the molecule. When a molecule contains a specific structural feature, the corresponding bit in the fingerprint is set to 1, and if not, the corresponding bit in the fingerprint is set to 0.

[0062] Step S2, determining the toxicity level of each characteristic compound in the fingerprint characteristic matrix based on the chemical substance toxicity judgment standard, and clustering and classifying each characteristic compound based on the toxicity level of each characteristic compound to obtain multiple toxicity category sets; wherein one toxicity category set corresponds to one toxicity level.

[0063] Specifically, the toxicity level of each characteristic compound in the fingerprint characteristic matrix is ​​determined with reference to the chemical substance toxicity judgment standard formulated by the environmental protection department or relevant international organizations.

[0064] Step S3: determining the information gain corresponding to each fingerprint position in the fingerprint feature matrix based on each of the toxicity category sets.

[0065] Specifically, step S3 includes the following steps:

[0066] Calculate the conditional probabilities corresponding to each of the toxicity category sets respectively, and obtain the overall entropy based on the conditional probabilities corresponding to each of the toxicity category sets; specifically, for each of the toxicity category sets, the conditional probability corresponding to the toxicity category set is the ratio of the number of molecules in the toxicity category set to the number of molecules in the molecular characteristic data information, and use formula (1) to obtain the overall entropy;

[0067] ;

[0068] in, H(T) is the overall entropy; For the The conditional probability corresponding to the set of toxicity categories; is the number of sets of toxicity categories;

[0069] Determine the conditional entropy corresponding to each of the fingerprint positions respectively; specifically, use formula (2) and formula (3) to determine the conditional entropy corresponding to each of the fingerprint positions respectively;

[0070] ;

[0071] ;

[0072] in, Fingerprint bit The corresponding conditional entropy; The fingerprint bit value is probability; It is the fingerprint bit is set to The individual entropy corresponding to the time; For the The molecules corresponding to the toxicity category set are in the fingerprint is set to The probability of existence in, for example, fingerprint The number of 1s is 100, The molecules corresponding to the toxicity category set are in the fingerprint is set to 1 and the number of is 0.2;

[0073] For each of the fingerprint bits, the conditional entropy corresponding to the fingerprint bit is subtracted from the overall entropy to obtain the information gain corresponding to the fingerprint bit.

[0074] Step S4: generating an initial key feature matrix based on the information gain corresponding to each of the fingerprint positions.

[0075] Specifically, for each fingerprint bit, the information gain corresponding to the fingerprint bit is compared with a preset information gain. If the information gain corresponding to the fingerprint bit is less than the preset information gain, the column vector corresponding to the fingerprint bit is deleted from the fingerprint feature matrix.

[0076] It can be understood that step S4 can enable the subsequent integrated biological toxicity prediction model to focus on the features that are most helpful in distinguishing different toxicity levels, thereby improving the accuracy of prediction by deleting the column vectors corresponding to the fingerprint positions whose information gain is less than the preset information gain in the fingerprint feature matrix.

[0077] Step S5, performing dimensionality reduction processing on the initial key feature matrix to obtain a key feature matrix;

[0078] Specifically, step S5 includes the following steps:

[0079] Perform binary combination on each initial key fingerprint position in the initial key feature matrix to obtain a plurality of binary combination pairs; it can be understood that the plurality of binary combination pairs are all possible combination pairs of the initial key fingerprint positions;

[0080] For each of the binary combination pairs, calculating the Pearson correlation coefficient between the two initial key fingerprint positions of the binary combination pair;

[0081] For each of the binary combination pairs, if the Pearson correlation coefficient corresponding to the binary combination pair is greater than a preset threshold, the column vector corresponding to the initial key fingerprint position with smaller information gain in the binary combination pair is deleted from the initial key feature matrix.

[0082] It can be understood that step S5 helps to improve the signal-to-noise ratio of the key feature matrix by performing dimensionality reduction processing on the initial key feature matrix, further improve the accuracy of wastewater biological toxicity prediction, and at the same time, reduce the computational complexity of the subsequent integrated biological toxicity prediction model.

[0083] Step S6: input the key feature matrix into a preset integrated biological toxicity prediction model to obtain a biological toxicity prediction result of the wastewater to be tested.

[0084] The method provided in this embodiment, on the one hand, determines the information gain corresponding to each fingerprint position in the fingerprint feature matrix based on each toxicity category set; generates an initial key feature matrix based on the information gain corresponding to each fingerprint position; performs dimensionality reduction processing on the initial key feature matrix to obtain a key feature matrix, thereby achieving effective feature screening of molecular feature data information, which helps to improve the accuracy of wastewater biological toxicity prediction results. On the other hand, by inputting the key feature matrix into a preset integrated biological toxicity prediction model, the biological toxicity prediction result of the wastewater to be tested is obtained, thereby further improving the accuracy of the wastewater biological toxicity prediction result.

[0085] In some embodiments, the training method of the integrated biological toxicity prediction model includes:

[0086] Acquire a training data set; wherein the training data set includes a plurality of samples, each sample includes a feature matrix and a corresponding biological toxicity standard value; wherein the feature matrix is ​​obtained by effectively extracting features from molecular feature data information of wastewater;

[0087] Dividing the training data set into a training set, a validation set and a test set;

[0088] Based on the training set, each preset basic learner is trained respectively to obtain a sub-model corresponding to each basic learner; wherein the basic learners include a linear regression model, a logistic regression model, a support vector machine, a convolutional neural network, a recursive neural network, etc.;

[0089] For each of the sub-models, judging whether the model parameters of the sub-model need to be adjusted based on the validation set; if necessary, adjusting the model parameters of the sub-model based on a preset parameter adjustment method to obtain a target sub-model corresponding to the sub-model; if not necessary, determining the sub-model as the target sub-model;

[0090] Determine the weight coefficient of each target sub-model based on the test set;

[0091] The integrated biological toxicity prediction model is constructed based on the weight coefficients of each target sub-model.

[0092] The step of determining whether the model parameters of the sub-model need to be adjusted based on the validation set comprises the following steps:

[0093] For each verification sample in the verification set, input the feature matrix corresponding to the verification sample into the sub-model to obtain the biological toxicity prediction value corresponding to the verification sample, and obtain the first absolute value of the difference between the biological toxicity standard value corresponding to the verification sample and the biological toxicity prediction value;

[0094] respectively comparing each of the first absolute values ​​with a first preset absolute value;

[0095] If any of the first absolute values ​​is greater than the first preset absolute value, it is determined that the model parameters of the sub-model need to be adjusted.

[0096] The method of adjusting the model parameters of the sub-model based on a preset parameter adjustment method to obtain a target sub-model corresponding to the sub-model includes the following steps:

[0097] For each of the verification samples, when the first absolute value corresponding to the verification sample is greater than the first preset absolute value, a binary cross entropy loss function value is constructed based on the biological toxicity standard value and the biological toxicity prediction value corresponding to the verification sample, and the output error value is obtained by subtracting the biological toxicity prediction value from the biological toxicity standard value, and based on the binary cross entropy loss function value and the output error value, the model parameters of the sub-model are adjusted using the back propagation method to obtain the target sub-model corresponding to the sub-model.

[0098] The step of determining the weight coefficients of the target sub-models based on the test set comprises the following steps:

[0099] For each of the target sub-models, determining the prediction accuracy of the target sub-model based on the test set;

[0100] Adding the prediction accuracy of each target sub-model to obtain the sum of the prediction accuracy;

[0101] For each of the target sub-models, the ratio of the prediction accuracy rate corresponding to the target sub-model to the sum of the prediction accuracy rates is used as the weight coefficient of the target sub-model.

[0102] Wherein, determining the prediction accuracy of the target sub-model based on the test set comprises the following steps:

[0103] For each test sample in the test set, input the feature matrix corresponding to the test sample into the target sub-model to obtain the biological toxicity prediction value corresponding to the test sample, and obtain the second absolute value of the difference between the biological toxicity standard value corresponding to the test sample and the biological toxicity prediction value;

[0104] Compare each of the second absolute values ​​with a second preset absolute value respectively; wherein the second preset absolute value is smaller than the first preset absolute value;

[0105] If the second absolute value is not greater than the second preset absolute value, determining that the test sample corresponding to the second absolute value is an accurate test sample;

[0106] The number of accurate test samples is counted, and the ratio between the number of accurate test samples and the total number of samples in the test set is determined as the prediction accuracy of the target sub-model.

[0107] The method provided in this embodiment, on the one hand, uses the integrated biological toxicity prediction model obtained by the method provided in this embodiment to help reduce the prediction error of a single model and improve the prediction accuracy of wastewater biological toxicity. On the other hand, by reasonably allocating weights to each of the sub-models, it helps to give full play to the advantages of high-performance models and improve the prediction accuracy of the integrated biological toxicity prediction model.

[0108] See also Figure 2 , Figure 2 A schematic block diagram of the structure of the wastewater biological toxicity prediction system 100 based on AI provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the AI-based wastewater biological toxicity prediction system 100 provided in the embodiment of the present application includes:

[0109] The acquisition module 110 is used to acquire molecular characteristic data information of the wastewater to be tested, and generate a fingerprint characteristic matrix corresponding to the molecular characteristic data information based on a preset molecular fingerprint algorithm.

[0110] The clustering classification module 120 is used to determine the toxicity level of each characteristic compound in the fingerprint characteristic matrix based on the chemical substance toxicity judgment standard, and cluster and classify each characteristic compound based on the toxicity level of each characteristic compound to obtain multiple toxicity category sets; wherein one toxicity category set corresponds to one toxicity level.

[0111] The determination module 130 is used to determine the information gain corresponding to each fingerprint bit in the fingerprint feature matrix based on each of the toxicity category sets.

[0112] The generating module 140 is configured to generate an initial key feature matrix based on the information gain corresponding to each of the fingerprint bits.

[0113] The dimension reduction processing module 150 is used to perform dimension reduction processing on the initial key feature matrix to obtain a key feature matrix.

[0114] The input module 160 is used to input the key feature matrix into a preset integrated biological toxicity prediction model to obtain the biological toxicity prediction result of the wastewater to be tested.

[0115] It should be noted that technicians in the relevant technical field can clearly understand that for the convenience and conciseness of description, the specific working process of the system and each module described above can refer to the corresponding process in the aforementioned AI-based wastewater biological toxicity prediction method embodiment, and will not be repeated here.

[0116] The AI-based wastewater biological toxicity prediction system 100 provided in the above embodiment can be implemented in the form of a computer program. The computer program can be used in Figure 3 The system is run on the terminal device 200 shown.

[0117] See also Figure 3 , Figure 3 The present invention provides a schematic block diagram of the structure of a terminal device 200 according to an embodiment of the present application. The terminal device 200 includes a processor 201 and a memory 202. The processor 201 and the memory 202 are connected via a device bus 203, wherein the memory 202 may include a non-volatile storage medium and an internal memory.

[0118] The non-volatile storage medium can store a computer program. The computer program includes program instructions, and when the program instructions are executed by the processor 201, the processor 201 can execute any of the above-mentioned AI-based wastewater biological toxicity prediction methods.

[0119] The processor 201 is used to provide computing and control capabilities to support the operation of the entire terminal device 200.

[0120] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor 201, the processor 201 can execute any of the above-mentioned AI-based wastewater biological toxicity prediction methods.

[0121] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the terminal device 200 involved in the scheme of the present application. The specific terminal device 200 may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0122] It should be understood that the processor 201 may be a central processing unit (CPU), and the processor 201 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0123] In some embodiments, the processor 201 is used to run a computer program stored in the memory to implement the following steps:

[0124] Acquire molecular characteristic data information of the wastewater to be tested, and generate a fingerprint characteristic matrix corresponding to the molecular characteristic data information based on a preset molecular fingerprint algorithm;

[0125] Determine the toxicity level of each characteristic compound in the fingerprint characteristic matrix based on the chemical substance toxicity judgment standard, and cluster and classify each characteristic compound based on the toxicity level of each characteristic compound to obtain multiple toxicity category sets; wherein one toxicity category set corresponds to one toxicity level;

[0126] Determine the information gain corresponding to each fingerprint bit in the fingerprint feature matrix based on each of the toxicity category sets;

[0127] Generate an initial key feature matrix based on the information gain corresponding to each of the fingerprint positions;

[0128] Performing dimensionality reduction processing on the initial key feature matrix to obtain a key feature matrix;

[0129] The key feature matrix is ​​input into a preset integrated biological toxicity prediction model to obtain a biological toxicity prediction result of the wastewater to be tested.

[0130] It should be noted that technicians in the relevant field can clearly understand that for the convenience and simplicity of description, the specific working process of the terminal device 200 described above can refer to the corresponding process of the aforementioned AI-based wastewater biological toxicity prediction method, and will not be repeated here.

[0131] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by one or more processors, the one or more processors implement the AI-based wastewater biological toxicity prediction method provided in the embodiment of the present application.

[0132] The computer-readable storage medium may be an internal storage unit of the terminal device 200 in the aforementioned embodiment, such as a hard disk or memory of the terminal device 200. The computer-readable storage medium may also be an external storage device of the terminal device 200, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped with the terminal device 200.

[0133] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A wastewater biological toxicity prediction method based on AI, characterized in that: include: Acquire molecular characteristic data information of the wastewater to be tested, and generate a fingerprint characteristic matrix corresponding to the molecular characteristic data information based on a preset molecular fingerprint algorithm; Determine the toxicity level of each characteristic compound in the fingerprint characteristic matrix based on the chemical substance toxicity judgment standard, and cluster and classify each characteristic compound based on the toxicity level of each characteristic compound to obtain multiple toxicity category sets; wherein one toxicity category set corresponds to one toxicity level; Determine the information gain corresponding to each fingerprint bit in the fingerprint feature matrix based on each of the toxicity category sets; Generate an initial key feature matrix based on the information gain corresponding to each of the fingerprint positions; Performing dimensionality reduction processing on the initial key feature matrix to obtain a key feature matrix; The key feature matrix is ​​input into a preset integrated biological toxicity prediction model to obtain a biological toxicity prediction result of the wastewater to be tested.

2. The AI-based wastewater biological toxicity prediction method according to claim 1 is characterized in that: The determining the information gain corresponding to each fingerprint bit in the fingerprint feature matrix based on each of the toxicity category sets respectively includes: Calculating the conditional probabilities corresponding to each of the toxicity category sets respectively, and obtaining the overall entropy based on the conditional probabilities corresponding to each of the toxicity category sets; Determine the conditional entropy corresponding to each of the fingerprint positions respectively; For each of the fingerprint bits, the conditional entropy corresponding to the fingerprint bit is subtracted from the overall entropy to obtain the information gain corresponding to the fingerprint bit.

3. The AI-based wastewater biological toxicity prediction method according to claim 1 is characterized in that: The step of performing dimensionality reduction processing on the initial key feature matrix to obtain the key feature matrix includes: Performing binary combination on each initial key fingerprint position in the initial key feature matrix to obtain a plurality of binary combination pairs; For each of the binary combination pairs, calculating the Pearson correlation coefficient between the two initial key fingerprint positions of the binary combination pair; For each of the binary combination pairs, if the Pearson correlation coefficient corresponding to the binary combination pair is greater than a preset threshold, the column vector corresponding to the initial key fingerprint position with smaller information gain in the binary combination pair is deleted from the initial key feature matrix.

4. The AI-based wastewater biological toxicity prediction method according to claim 1, characterized in that: The training method of the integrated biological toxicity prediction model comprises: Obtaining a training data set; wherein the training data set includes a plurality of samples, each of which includes a feature matrix and a corresponding biological toxicity standard value; Dividing the training data set into a training set, a validation set and a test set; Based on the training set, each preset basic learner is trained respectively to obtain a sub-model corresponding to each basic learner; For each of the sub-models, judging whether the model parameters of the sub-model need to be adjusted based on the validation set; if necessary, adjusting the model parameters of the sub-model based on a preset parameter adjustment method to obtain a target sub-model corresponding to the sub-model; if not necessary, determining the sub-model as the target sub-model; Determine the weight coefficient of each target sub-model based on the test set; The integrated biological toxicity prediction model is constructed based on the weight coefficients of each target sub-model.

5. The AI-based wastewater biological toxicity prediction method according to claim 4 is characterized in that: The determining whether the model parameters of the sub-model need to be adjusted based on the validation set includes: For each verification sample in the verification set, input the feature matrix corresponding to the verification sample into the sub-model to obtain the biological toxicity prediction value corresponding to the verification sample, and obtain the first absolute value of the difference between the biological toxicity standard value corresponding to the verification sample and the biological toxicity prediction value; respectively comparing each of the first absolute values ​​with a first preset absolute value; If any of the first absolute values ​​is greater than the first preset absolute value, it is determined that the model parameters of the sub-model need to be adjusted.

6. The AI-based wastewater biological toxicity prediction method according to claim 5 is characterized in that: The method of adjusting the model parameters of the sub-model based on a preset parameter adjustment method to obtain a target sub-model corresponding to the sub-model includes: For each of the verification samples, when the first absolute value corresponding to the verification sample is greater than the first preset absolute value, a binary cross entropy loss function value is constructed based on the biological toxicity standard value and the biological toxicity prediction value corresponding to the verification sample, and the output error value is obtained by subtracting the biological toxicity prediction value from the biological toxicity standard value, and based on the binary cross entropy loss function value and the output error value, the model parameters of the sub-model are adjusted using the back propagation method to obtain the target sub-model corresponding to the sub-model.

7. The AI-based wastewater biological toxicity prediction method according to claim 6, characterized in that: The step of determining the weight coefficients of the target sub-models based on the test set includes: For each of the target sub-models, determining the prediction accuracy of the target sub-model based on the test set; Adding the prediction accuracy of each target sub-model to obtain the sum of the prediction accuracy; For each of the target sub-models, the ratio of the prediction accuracy rate corresponding to the target sub-model to the sum of the prediction accuracy rates is used as the weight coefficient of the target sub-model.

8. The AI-based wastewater biological toxicity prediction method according to claim 7, characterized in that: The determining the prediction accuracy of the target sub-model based on the test set includes: For each test sample in the test set, input the feature matrix corresponding to the test sample into the target sub-model to obtain the biological toxicity prediction value corresponding to the test sample, and obtain the second absolute value of the difference between the biological toxicity standard value corresponding to the test sample and the biological toxicity prediction value; Compare each of the second absolute values ​​with a second preset absolute value respectively; wherein the second preset absolute value is smaller than the first preset absolute value; If the second absolute value is not greater than the second preset absolute value, determining that the test sample corresponding to the second absolute value is an accurate test sample; The number of accurate test samples is counted, and the ratio between the number of accurate test samples and the total number of samples in the test set is determined as the prediction accuracy of the target sub-model.

9. An AI-based wastewater biological toxicity prediction system, characterized in that: include: An acquisition module is used to acquire molecular characteristic data information of the wastewater to be tested, and generate a fingerprint characteristic matrix corresponding to the molecular characteristic data information based on a preset molecular fingerprint algorithm; A clustering classification module, used to determine the toxicity level of each characteristic compound in the fingerprint characteristic matrix based on the chemical substance toxicity judgment standard, and cluster and classify each characteristic compound based on the toxicity level of each characteristic compound to obtain multiple toxicity category sets; wherein one toxicity category set corresponds to one toxicity level; A determination module, configured to determine the information gain corresponding to each fingerprint position in the fingerprint feature matrix based on each of the toxicity category sets; A generating module, used for generating an initial key feature matrix based on the information gain corresponding to each of the fingerprint positions; A dimension reduction processing module is used to perform dimension reduction processing on the initial key feature matrix to obtain a key feature matrix; The input module is used to input the key feature matrix into a preset integrated biological toxicity prediction model to obtain the biological toxicity prediction result of the wastewater to be tested.

Images