Prediction method and system for post-treatment amines of organic-inorganic hybrid perovskite based on random forest fast classification

Abstract

The invention discloses a prediction method and system for organic-inorganic hybrid perovskite post-treatment amines based on random forest rapid classification, and the specific contents include: establishing a data set sample; generating descriptors; randomly dividing a training set and a test set; screening variables using a maximum correlation minimum redundancy method to select an optimal variable subset for modeling; establishing a classification model for rapid classification of organic-inorganic hybrid perovskite post-treatment amines using a random forest classification algorithm; and rapidly predicting the categories of 10 amines of an independent test set based on the established rapid classification model of amines. The invention is based on reliable literature true values ​​and modeling methods, and the classification model for organic-inorganic hybrid perovskite post-treatment amines established has the advantages of being convenient and fast, and having no chemical pollution.

Description

Technical Field

[0001] The invention relates to a method and system for classifying post-treatment amines of organic-inorganic hybrid perovskite, which is a method for fast classification of post-treatment amines based on random forests. The method is applied to improving the stability of organic-inorganic hybrid perovskite solar cells. Background Art

[0002] Perovskite refers to a class of ceramic oxides with the general molecular formula ABO 3 , the A-site ion is usually a rare earth or alkaline earth metal element with a large ion radius, and the B-site is generally an element with a smaller ion radius, generally a transition metal element, such as Mn, Co, Fe, etc. Perovskite has a huge development prospect in the field of optoelectronics due to its specific stable crystal structure, so perovskite solar cells have become a research hotspot in recent years. Perovskite materials used in solar cells can not only achieve a higher photoelectric conversion efficiency than other solar cells, but also have a relatively simple manufacturing process, low price, and good stability.

[0003] Organic-inorganic hybrid perovskite (HOIPs) materials are a new type of composite crystal materials composed of organic and inorganic molecules. Inorganic components can provide thermal stability, hardness, magnetic and dielectric properties for the hybrid structure; organic ligands can provide structural diversity and efficient electrical and optical properties. In recent years, organic-inorganic hybrid perovskites have been well developed and applied in the field of solar cells.

[0004] Traditional solar cells are made of silicon in the active layer, and it took decades of improvements for these devices to reach an efficiency of more than 20%. It took perovskite solar cells only 10 years to reach the same level. But perovskite solar cells have a very obvious problem, which is their poor stability. Various studies have found that perovskites have poor stability under conditions such as moisture, oxygen, and ultraviolet rays. Therefore, post-processing perovskites to improve their stability is of great research significance for the future development of solar cells.

[0005] Random Forest (RF) is a very representative bagging ensemble algorithm that uses random samples and random features to construct a single base estimator. All of its base estimators are decision trees, and can be adapted to classification and regression tasks at the same time depending on the decision tree. The advantages of the random forest algorithm are: (1) it can produce highly accurate classifiers; (2) it can handle large data sets; (3) it uses unbiased estimation in the modeling process; (4) it can handle input samples with high-dimensional features without the need for dimensionality reduction; (5) it can evaluate the importance of each feature in the classification problem, etc.

[0006] Maximum relevance minimum redundancy (mRMR) is a common method for independent variable screening. This method is a filtering feature screening algorithm that screens independent variables based on the theory that the redundancy between the selected features should be minimal and the correlation with the target variable should be maximal. However, how to apply the random forest algorithm to quickly predict the amines of post-treatment of organic-inorganic hybrid perovskites requires further exploration and research. Summary of the invention

[0007] In order to solve the problems of the prior art, the purpose of the present invention is to overcome the shortcomings of the prior art and provide a prediction method and system for organic-inorganic hybrid perovskite post-treatment amines based on random forest rapid classification, which can quickly and accurately classify the amines used for organic-inorganic hybrid perovskite post-treatment, thereby improving the problem of poor stability of perovskite solar cells. The method calculates the descriptors of the amines, uses the maximum correlation minimum redundancy to screen variables, and predicts the category of the amines through random forest algorithm modeling. The method can obtain classification results within a few minutes, which is convenient and fast, without the need for experiments and complicated calculations.

[0008] In order to achieve the above invention purpose, the present invention adopts the following technical solution:

[0009] A prediction method for post-treatment amines of organic-inorganic hybrid perovskite based on random forest rapid classification comprises the following steps:

[0010] (1) Establishing a data set sample: Using a computer system, search the literature for the structures of amines used for post-treatment of organic-inorganic hybrid perovskites as data set samples;

[0011] (2) Generate descriptors: Use Chemdraw to draw the structures of the collected amines, and then use Dragon software to generate descriptors;

[0012] (3) Taking whether the perovskite film reacts with the post-treatment amine as the target variable, the descriptors generated by Dragon were used as independent variables, and the variable autocorrelation was 0.95 to preliminarily screen the characteristic variables;

[0013] (4) Use the maximum relevance and minimum redundancy method to further screen the variables and select the optimal variable subset for modeling;

[0014] (5) Randomly divide the training set and test set, and the proportion of the test set accounts for 20% of the entire data set;

[0015] (6) A classification model for post-treatment amines was established using a random forest classification algorithm;

[0016] (7) Based on the established model, the structural formulas of at least 10 amines in an independent test set are quickly predicted.

[0017] Preferably, in the step (4), the method of further screening the characteristic variables using the maximum relevance minimum redundancy method is as follows:

[0018] Using HyperMiner data mining software, import the initially screened data set, select the maximum correlation and minimum redundancy method, determine the target variable and independent variable, click "Calculate", obtain the relative importance ranking of the feature variables, and on this basis select the feature variables for modeling.

[0019] A system for predicting post-treatment amines for organic-inorganic hybrid perovskites, performing a prediction method for post-treatment amines for organic-inorganic hybrid perovskites based on random forest fast classification of the present invention, comprising:

[0020] Input module: Search for structural data of amines used for post-treatment of organic-inorganic hybrid perovskites from publicly available literature and use them as input data;

[0021] Data analysis module: using the data obtained from the input module, executing a prediction method based on random forest fast classification for post-treatment amines of organic-inorganic hybrid perovskite, and quickly predicting the post-treatment amines of organic-inorganic hybrid perovskite;

[0022] Output module: Output the amine data of post-processing of organic-inorganic hybrid perovskite.

[0023] Compared with the prior art, the present invention has the following obvious outstanding substantive features and significant advantages:

[0024] 1. The random forest rapid classification method of the present invention avoids repeated experiments, overcomes the shortcomings of the traditional experimental "trial and error method", uses Dragon software to generate descriptors for the drawn amine structure, and after variable screening and modeling, the data is imported into the model, and classification can be completed within seconds, which can provide guidance and reference for experimenters, thereby shortening the research and development time and reducing the research and development cost;

[0025] 2. The present invention generates independent variables in Dragon software and performs certain feature screening, and then uses random forest modeling. The operation process is simple, the cost is low, and manpower and resources are saved;

[0026] 3. The entire process of the present invention does not involve experiments and chemicals, does not cause environmental pollution, and is in line with the concept of green environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flowchart of the present invention.

[0028] Figure 2 This is a confusion matrix diagram of the modeling results of the random forest classification model for rapid classification of amines of the present invention.

[0029] Figure 3 This is a confusion matrix diagram of the leave-one-out cross-validation results of the random forest rapid classification amine classification model of the present invention. DETAILED DESCRIPTION

[0030] The above scheme is further described below in conjunction with specific implementation examples. The preferred embodiments of the present invention are described in detail as follows:

[0031] Embodiment 1:

[0032] In this embodiment, see Figure 1 ,,A prediction method for post-treatment amines of organic-inorganic hybrid perovskite based on random forest fast classification, comprising the following steps:

[0033] (1) Establishing a data set sample: Using a computer system, search the literature for the structures of amines used for post-treatment of organic-inorganic hybrid perovskites as data set samples;

[0034] (2) Generate descriptors: Use Chemdraw to draw the structures of the collected amines, and then use Dragon software to generate descriptors;

[0035] (3) Taking whether the perovskite film reacts with the post-treatment amine as the target variable, the descriptors generated by Dragon were used as independent variables, and the variable autocorrelation was 0.95 to preliminarily screen the characteristic variables;

[0036] (4) Use the maximum relevance and minimum redundancy method to further screen the variables and select the optimal variable subset for modeling;

[0037] (5) Randomly divide the training set and test set, and the proportion of the test set accounts for 20% of the entire data set;

[0038] (6) A classification model for post-treatment amines was established using a random forest classification algorithm;

[0039] (7) Based on the established model, the structural formulas of at least 10 amines in an independent test set are quickly predicted.

[0040] The random forest classification algorithm in step (6) of the method of this embodiment is an algorithm based on multiple decision trees, and has higher classification accuracy than the support vector machine classification algorithm used in previous related studies. This embodiment calculates the descriptors of the two-dimensional structure of amine molecules, uses the maximum correlation minimum redundancy to screen variables, and uses the random forest classification algorithm to model and predict the category of the predicted amine. These methods can obtain results in a few minutes, which is convenient and fast, without the need for experiments and complicated calculations.

[0041] Embodiment 2

[0042] This embodiment is basically the same as the first embodiment, except that:

[0043] In this embodiment, in step (4), the method of further screening feature variables using the maximum relevance minimum redundancy method is as follows:

[0044] Using HyperMiner data mining software, import the initially screened data set, select the maximum correlation and minimum redundancy method, determine the target variable and independent variable, click "Calculate", obtain the relative importance ranking of the feature variables, and on this basis select the feature variables for modeling.

[0045] This embodiment uses the maximum relevance minimum redundancy method to screen feature variables, and can quickly and accurately obtain the relative importance ranking of feature variables, which is convenient for subsequent modeling work. The method is simple and fast to operate, and hundreds or thousands of feature variables can be ranked in just a few seconds.

[0046] Embodiment 3

[0047] This embodiment is basically the same as the above embodiment, with the following features:

[0048] In this embodiment, a method for rapid classification of amines for post-treatment of organic-inorganic hybrid perovskite based on random forests includes the following steps:

[0049] (1) Using a computer system, we searched the literature for amines used for post-treatment of organic-inorganic hybrid perovskites as data set samples for machine learning. A total of 50 amines that met the requirements were found, and some of their structures are shown in Table 1:

[0050] Table 1. Structure of some amines

[0051]

[0052]

[0053] (2) A total of 5270 descriptors were generated for the two-dimensional structure of amines drawn by Chemdraw using Dragon software. Some of the descriptors are shown in Table 2:

[0054] Table 2. Descriptors of some amines generated by Dragon

[0055]

[0056]

[0057] (3) Whether the perovskite film reacts with the post-treatment amine is taken as the target variable, the amine that reacts is marked as category 0, and the amine that does not react is marked as category 1. The descriptor generated by Dragon is the independent variable, and 996 feature variables are initially screened out with a variable autocorrelation of 0.95;

[0058] (4) The feature variables were screened by maximum correlation and minimum correlation redundancy, and 10 optimal descriptors were selected, which are: P_VSA_ppp_D; nHDon; CATS2D_00_DD; Hy; ChiA_B(m); CIC4; SpMAD_AEA(dm); SpMAD_L; TPSA(NO); GATS2m. The data of the 10 optimal descriptors of some samples are shown in Table 3:

[0059] Table 3. Partial data of the 10 best descriptors

[0060]

[0061] 5) Randomly divide the training set and test set into 80%:20% groups, with the sample sizes of the training set and test set being 40 and 10 respectively;

[0062] 6) After the above data processing, a rapid classification model of post-processed amines was established using a random forest classification algorithm;

[0063] 7) Based on the established rapid classification model for post-treated amines in organic-inorganic hybrid perovskites and the amines to be classified, the categories of the amines in the independent test set are quickly predicted.

[0064] In this embodiment, the confusion matrix of the modeling effect of the amine classification model established based on the random forest classification algorithm is as follows: Figure 2 As shown in the figure, the random forest classification algorithm was used to classify and model the 40 amine sample data, and the classification accuracy reached 97.50%.

[0065] In this embodiment, the confusion matrix of the leave-one-out cross-validation result of the amine classification model established based on the random forest classification algorithm is as follows: Figure 3 The classification model established by the leave-one-out cross-validation method was cross-validated on 40 sample data, and the classification accuracy reached 80.00%.

[0066] In this embodiment, the independent test set prediction results of the amine classification model established based on the random forest classification algorithm are shown in Table 4.

[0067] Table 4. Classification results of the independent test set of the random forest rapid classification amine classification model of this embodiment

[0068]

[0069] As shown in Table 4, by establishing a prediction model of the random forest classification algorithm to predict the 10 samples in the independent test set, the prediction results are good and all 10 samples are classified correctly.

[0070] Embodiment 4

[0071] This embodiment is basically the same as the above embodiment, with the following features:

[0072] In this embodiment, a system for predicting post-treatment amines for organic-inorganic hybrid perovskites is implemented, and the prediction method for post-treatment amines for organic-inorganic hybrid perovskites based on random forest fast classification of the above embodiment is implemented, including:

[0073] Input module: Search for structural data of amines used for post-treatment of organic-inorganic hybrid perovskites from publicly available literature and use them as input data;

[0074] Data analysis module: using the data obtained from the input module, executing a prediction method based on random forest fast classification for post-treatment amines of organic-inorganic hybrid perovskite, and quickly predicting the post-treatment amines of organic-inorganic hybrid perovskite;

[0075] Output module: Output the amine data of post-processing of organic-inorganic hybrid perovskite.

[0076] The program of the system of this embodiment is based on the method of rapid classification of organic-inorganic hybrid perovskite post-treatment amines by random forests, establishes a data set sample; generates descriptors; randomly divides the training set and the test set; uses the maximum correlation minimum redundancy method to screen variables and select the optimal variable subset for modeling; uses the random forest classification algorithm to establish a classification model for rapid classification of organic-inorganic hybrid perovskite post-treatment amines; according to the established rapid classification model of amines, quickly predicts the categories of 10 amines in the independent test set. The system of this embodiment is based on reliable literature true values ​​and modeling methods, and the classification model of organic-inorganic hybrid perovskite post-treatment amines established has the advantages of being convenient and fast, and having no chemical pollution.

[0077] In summary, the above embodiment is based on the method for rapid classification of organic-inorganic hybrid perovskite post-treatment amines, which includes the following steps: (1) using a computer system to search for amines used for post-treatment of organic-inorganic hybrid perovskites from the literature as a data set sample; (2) using Chemdraw to draw the structure of the collected amines, and then using Dragon software to generate descriptors; (3) using whether the perovskite film reacts with the post-treatment amine as the target variable, the descriptor generated by Dragon as the independent variable, and preliminarily screening the characteristic variables with the variable autocorrelation of 0.95; (4) using the maximum correlation minimum redundancy method to further screen the variables and select the optimal variable subset for modeling; (5) randomly dividing the training set and the test set, with the proportion of the test set accounting for 20% of the entire data set; (6) using the random forest classification algorithm to establish a classification model for post-treatment amines; (7) based on the established model, quickly predicting 10 amines in the independent test set.

[0078] The above-mentioned embodiment is based on reliable literature real values ​​and modeling methods, and the prediction model of the inhibitory activity of xanthine derivatives established has the advantages of being convenient, fast, and free of chemical pollution.

[0079] The above embodiment method avoids repeated experiments and the process of trial and error. It uses Dragon software to generate descriptors for the drawn amine structure. After variable screening and random forest classification modeling, the amines can be quickly and accurately classified, which can provide a reference for experimenters, shorten research and development time, and reduce research and development costs. The above embodiment method is based on the generation of independent variables by Dragon software and certain screening, and then modeling with the random forest classification algorithm. The operation process is simple and low-cost, and can be completed by only one person.

[0080] The above describes the embodiments of the present invention in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Various changes can be made according to the purpose of the invention. Any changes, modifications, substitutions, combinations or simplifications made according to the spirit and principle of the technical solution of the present invention should be equivalent replacement methods. As long as they meet the purpose of the invention and do not deviate from the technical principles and inventive concepts of the present invention, they belong to the protection scope of the present invention.

Claims

1. A prediction method for post-treatment amines of organic-inorganic hybrid perovskite based on random forest rapid classification, characterized in that: The steps include: (1) Establishing a data set sample: Using a computer system, search the literature for the structures of amines used for post-treatment of organic-inorganic hybrid perovskites as data set samples; (2) Generate descriptors: Use Chemdraw to draw the structures of the collected amines, and then use Dragon software to generate descriptors; (3) Taking whether the perovskite film reacts with the post-treatment amine as the target variable, the descriptors generated by Dragon were used as independent variables, and the variable autocorrelation was 0.95 to preliminarily screen the characteristic variables; (4) Use the maximum relevance and minimum redundancy method to further screen the variables and select the optimal variable subset for modeling; (5) Randomly divide the training set and test set, and the proportion of the test set accounts for 20% of the entire data set; (6) A classification model for post-treatment amines was established using a random forest classification algorithm; (7) Based on the established model, the structural formulas of at least 10 amines in an independent test set are quickly predicted.

2. The method for predicting post-treatment amines of organic-inorganic hybrid perovskite based on random forest rapid classification according to claim 1, characterized in that: In the step (4), the method of further screening the characteristic variables using the maximum correlation minimum redundancy method is as follows: Using HyperMiner data mining software, import the initially screened data set, select the maximum correlation and minimum redundancy method, determine the target variable and independent variable, click "Calculate", obtain the relative importance ranking of the feature variables, and on this basis select the feature variables for modeling.

3. A system for predicting post-treatment amines for organic-inorganic hybrid perovskites, executing the prediction method for post-treatment amines for organic-inorganic hybrid perovskites based on random forest fast classification as described in any one of claims 1-2, characterized in that: include: Input module: Search for structural data of amines used for post-treatment of organic-inorganic hybrid perovskites from publicly available literature and use them as input data; Data analysis module: using the data obtained from the input module, executing a prediction method based on random forest fast classification for post-treatment amines of organic-inorganic hybrid perovskite, and quickly predicting the post-treatment amines of organic-inorganic hybrid perovskite; Output module: Output the amine data of post-processing of organic-inorganic hybrid perovskite.

Images