A method for predicting the impact of microplastics on algae growth

Abstract

The present invention discloses a method for predicting the effect of microplastics on algae growth, comprising: obtaining algae data to be predicted; inputting the algae data to be predicted into an algae growth prediction model to obtain algae growth conditions, wherein the algae growth prediction model is trained by a training set, the training set is a data set of algae growth conditions and corresponding numerical labels, and the algae growth prediction model is constructed using a machine learning algorithm. The present invention solves the problems of long time and high cost in traditional experimental methods for investigating and statistically analyzing algae growth conditions, and the invention has a high prediction accuracy.

Description

Technical Field

[0001] The present invention belongs to the field of machine learning regression prediction, and in particular relates to a method for predicting the impact of microplastics on algae growth. Background Art

[0002] As the primary producer of the ecosystem, the growth of algae is related to the stability of the ecosystem. Therefore, it is crucial to pay attention to the growth of algae. The traditional method of monitoring the growth density of algae is manual regular sampling survey. However, this method requires a lot of manpower and material resources, especially large-scale surveys are more difficult to implement. In recent years, with the global concern about microplastic pollution, more and more studies have focused on the impact of microplastics on the growth of algae. These studies often focus on the concentration, size and type of microplastics. The research content is very rich, covering various microplastic concentrations, types and sizes. However, these studies are also very scattered, and lack a systematic method to organize these research data.

[0003] In response to the above problems, the existing method can solve the long-term dependency problem between historical data and achieve accurate growth prediction of a certain algae. However, this method only focuses on one type of algae, ignoring the fact that there are many types of algae in the actual environment. The impact of microplastics on the growth of algae is very significant, but there is a lack of systematic methods to analyze the impact of microplastics on algae growth at home and abroad. Therefore, there is an urgent need for a prediction method for the impact of microplastics on algae growth. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a method for predicting the impact of microplastics on the growth of algae. This method can predict the growth conditions of algae in water bodies containing microplastics over a long period of time and over a large range, thereby improving work efficiency and making it easier to study the environmental benefits of microplastics. This has practical significance and good application prospects.

[0005] To achieve the above objectives, the present invention provides a method for predicting the effect of microplastics on algae growth, comprising:

[0006] Acquire algae data to be predicted;

[0007] The algae data to be predicted is input into an algae growth prediction model to obtain algae growth conditions, wherein the algae growth prediction model is trained by a training set, and the training set is a data set of algae growth conditions.

[0008] Optionally, the data set of algae growth conditions includes: microplastic size, microplastic concentration, microplastic type, algae type, algae initial density, culture medium, culture time, light intensity, photoperiod, culture temperature, culture speed and additives.

[0009] Optionally, before using the training set to train the algae growth prediction model, the method further includes: preprocessing the data set.

[0010] Optionally, preprocessing the data set includes: one-hot encoding the microplastic types, algae types, culture media and additives.

[0011] Optionally, the algae growth prediction model is a GBM model.

[0012] Optionally, the algae growth prediction model also includes using R 2 , MSE, RMSE, and MAE were used to verify the algae growth prediction model.

[0013] Optional,

[0014] The R 2 The calculation formula is:

[0015]

[0016] The calculation formula of the MSE is:

[0017]

[0018] The calculation formula of the RMSE is:

[0019]

[0020] The calculation formula of the MAE is:

[0021]

[0022] in, represents the true value, represents the mean of the true values, Represents the predicted value, and N represents the number of samples.

[0023] Compared with the prior art, the present invention has the following advantages and technical effects:

[0024] The present invention can predict the growth conditions of algae in water bodies containing microplastics over a long period of time and over a large range, thereby improving work efficiency and making it easier to study the environmental benefits of microplastics. This has practical significance and good application prospects.

[0025] The present invention utilizes machine learning regression technology to perform model construction, model training, model verification, model prediction and other steps on the algae growth status data set, and targets 12 independent variable factors existing in the experimental environment, and finally quickly predicts the growth inhibition of algae by the 12 factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0027] Figure 1 This is a flow chart of a method for predicting the effect of microplastics on algae growth according to an embodiment of the present invention;

[0028] Figure 2 A schematic diagram of a GBM model flow chart of an embodiment of the present invention;

[0029] Figure 3 is the R of multiple models in the embodiment of the present invention 2 and MSE value graph;

[0030] Figure 4 RMSE and MAE value diagrams of multiple models in embodiments of the present invention;

[0031] Figure 5 A comparison chart of the actual value and the predicted value of the growth inhibition rate of algae affected by microplastics predicted by the GBM model of an embodiment of the present invention;

[0032] Figure 6 SHAP value diagram of 12 independent variable factors of the embodiment of the present invention;

[0033] Figure 7 SHAP feature importance performance diagram of the growth inhibition condition of an embodiment of the present invention;

[0034] Figure 8 It is the SHAP value graph of a single sample in the data set. DETAILED DESCRIPTION

[0035] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0036] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0037] The present invention proposes a method for predicting the effect of microplastics on algae growth, which is characterized by comprising:

[0038] Acquire algae data to be predicted;

[0039] The algae data to be predicted is input into the algae growth prediction model to obtain the algae growth status, wherein the algae growth prediction model is trained by a training set, and the training set is a data set of the algae growth status.

[0040] Furthermore, the data set of algae growth conditions includes: microplastic size, microplastic concentration, microplastic type, algae type, algae initial density, culture medium, culture time, light intensity, photoperiod, culture temperature, culture speed and additives.

[0041] Specifically, the data set of algae growth conditions is 203 pieces of real experimental data under different environments, of which 185 pieces are from the prior art and 18 pieces are from the experimental data of this application. Each set of data represents 12 different independent variable factors.

[0042] The 203 sets of data were divided into training sets (162 sets) and validation sets (41 sets) in a ratio of 8:2 for building and evaluating machine learning models.

[0043] Furthermore, before using the training set to train the algae growth prediction model, the method also includes: preprocessing the data set.

[0044] Specifically, in order to eliminate the impact of data dimension and balance the weight of input features, the data was preprocessed before model training, and one-hot encoding was performed on all input category data features: microplastic types, algae types, culture media, and additives.

[0045] Furthermore, the algae growth prediction model is a GBM model. GBM generates multiple weak learners in series. Each round of iteration generates a weak learner. This weak learner fits the gradient of the loss function with respect to the previous cumulative model. Then, this weak learner is added to the cumulative model to gradually reduce the loss of the cumulative model. That is, the gradient descent in the parameter space uses the gradient information to adjust the parameters, thereby reducing the loss, while the gradient descent in the function space uses the gradient to fit a new function, thereby reducing the loss. It linearly combines the base learners with different weights so that the learners with excellent performance can be reused. The most commonly used learner is the tree model.

[0046] Specifically, GBM is a special "Boosting" algorithm, which attempts to sequentially iterate weak learners to learn from errors and develop a robust model, thereby achieving better learning results than a single weak learner. Figure 2The following is a schematic diagram of the GBM model flow. For each iteration, the model will try to minimize the loss function and learn from previous errors. The final model will be a collection of all weak learners, which summarizes all predictions of each model. Compared with models such as GLM and RF, GBM is insensitive to data with interference noise and can fit complex nonlinear relationships. These structural and algorithmic logic features enable GBM to build more accurate models and improve model generalization capabilities.

[0047] Furthermore, the algae growth prediction model also includes the use of R 2 , MSE, RMSE, and MAE were used to verify the algae growth prediction model.

[0048] Specifically, R 2 The MSE value indicates the degree of agreement between the predicted value and the actual value. The MSE indicator measures the squared average of the error or deviation. The RMSE indicator evaluates the ability of the model to predict continuous values. MAE is the average of the absolute errors.

[0049] Further,

[0050] R 2 The calculation formula is:

[0051]

[0052] The calculation formula of MSE is:

[0053]

[0054] The calculation formula of RMSE is:

[0055]

[0056] The calculation formula of MAE is:

[0057]

[0058] in, represents the true value, represents the mean of the true values, Represents the predicted value, and N represents the number of samples.

[0059] The present invention Figure 3 R showing three algorithms 2 and MSE values, where R 2 The larger the value, the closer the predicted result is to the actual result. 2The value reached 0.88, achieving significantly superior results in the modeling of data by the three algorithms. At the same time, in the comparison of MSE (the smaller the better), GLM is much more sensitive to input data features than RF and GBM, reaching 387.6, while GBM is less sensitive to input data features, only 65.8. In particular, Figure 4 The RMSE (the smaller the better) and MAE (the smaller the better) of the three algorithms are shown. GBM also achieved the smallest RMSE and MAE in the comparison of the three models. The excellent prediction performance of GBM reflects the structure and mechanism of the algorithm.

[0060] Figure 5 The GBM model fits the actual value and predicted value of the input data. The fit is very good. 2 The value reached 0.88, demonstrating strong model generalization ability and data feature extraction ability. The SHAP summary graph combines feature importance and feature effect by summarizing feature parameters and each SHAP value of the instance. In the present invention, the SHAP value is a characterization of the degree of influence of a given feature parameter on the growth inhibition of algae by microplastics. Figure 6The SHAP value graph of 12 independent variable factors is shown. The y-axis ranks the characteristic parameters according to the degree of their influence on the inhibition of algae growth, and the x-axis represents the SHAP value of each data point. The higher the absolute value of the x-axis, the stronger the impact on the model prediction. A positive SHAP value indicates a positive impact, and a negative value indicates a negative impact. The initial density of microplastics occupies the highest position on the y-axis, and the distribution of SHAP values ​​on the x-axis is the widest, indicating that the initial density of microplastics has an important impact on the growth trend of algae. At a higher initial density of microplastics, microplastics promote the growth of algae. On the contrary, at a lower initial density of microplastics, microplastics inhibit the growth of algae. This may be because the relative concentration of microplastics to algae is greater than that at a lower initial density of algae, which is not conducive to the growth of algae. Compared with a higher initial density of algae, the relative concentration of microplastics to algae is smaller. The factors of microplastic concentration and incubation time are also at a high position on the y-axis, and the distribution on the x-axis is also at a relatively wide position, indicating that the factors of microplastic concentration and incubation time also affect the growth of algae to a great extent. The higher the microplastic concentration, the more it inhibits the growth of algae, and the lower the microplastic concentration, the more it promotes the growth of algae. For the incubation time, the longer the incubation time, the better the algae grow, and microplastics can promote the growth of algae. The shorter the incubation time, the more obvious the inhibition of algae growth by microplastics. During the cultivation process, algae are stressed by high concentrations of microplastics, which inhibits the growth of algae. On the one hand, with the increase of incubation time, the dominant algae that adapt to microplastics grow, and can even use microplastics as a carbon source. On the other hand, with the increase of algae density, the concentration of microplastics is relatively low. These are all important factors that promote the growth of algae. The types of microplastics are at the lowest position on the y-axis and have the narrowest distribution on the x-axis, indicating that there is no obvious relationship between the types of microplastics and the model predictions. Figure 7The feature importance performance graph of the SHAP values ​​used to predict the inhibition of algae growth by microplastics during training is shown. Initial density of algae, incubation time, microplastic concentration, algae species, and microplastic size play a significant role (the light period and types of microplastics values ​​are close to 0, which is almost ineffective, Figure 8 No specific value is given).

[0061] After the constructed algae growth prediction model is trained, verified and saved, a set of data is randomly selected. Figure 8 As shown in the figure (SHAP value of a single sample in the data set, which shows the influence of various factors on the model prediction value). For this set of data, Prediction = 7.79% represents the model's predicted value of the growth inhibition rate of this sample. The initial algae concentration (Initial density of algae) is 100000 (cells / ml), and the SHAP value is -14, which has the greatest negative impact on the growth inhibition rate of algae. In other words, the higher the initial algae concentration (Initial density of algae), the more it promotes the growth of algae. Figure 6 The SHAP graph analysis results of all samples in the data set were consistent. At the same time, when the incubation time of this group of data was 96 (h), the microplastics concentration was 50 (mg / L), and the algae species was 110 (110 is a binary one-hot code, and 110 represents marine dinoflagellates in this study), the growth of marine dinoflagellates was inhibited.

[0062] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for predicting the effect of microplastics on algae growth, characterized in that: include: Acquire algae data to be predicted; Inputting the algae data to be predicted into an algae growth prediction model to obtain algae growth status, wherein the algae growth prediction model is trained by a training set, the training set is a data set of algae growth status and corresponding numerical labels, and the algae growth prediction model is constructed using a machine learning algorithm; The data set of algae growth conditions includes: microplastic size, microplastic concentration, microplastic type, algae type, algae initial density, culture medium, culture time, light intensity, photoperiod, culture temperature, culture speed and additives; Before using the training set to train the algae growth prediction model, the method further includes: preprocessing the data set; preprocessing the data set includes: one-hot encoding the microplastic species, algae species, culture medium and additives; The algae growth prediction model is a GBM model. GBM generates multiple weak learners in series. Each round of iteration generates a weak learner. The weak learner fits the gradient of the loss function with respect to the previous cumulative model. Then, the weak learner is added to the cumulative model to gradually reduce the loss of the cumulative model. That is, the gradient descent in the parameter space uses the gradient information to adjust the parameters, thereby reducing the loss, while the gradient descent in the function space uses the gradient to fit a new function, thereby reducing the loss. The base learners are linearly combined with different weights to reuse the learners with excellent performance.

2. The method for predicting the effect of microplastics on algae growth according to claim 1, characterized in that: The algae growth prediction model also includes using R 2 , MSE, RMSE, and MAE were used to verify the algae growth prediction model.

3. The method for predicting the effect of microplastics on algae growth according to claim 2, characterized in that: The R 2 The calculation formula is: The calculation formula of the MSE is: The calculation formula of the RMSE is: The calculation formula of the MAE is: in, represents the true value, represents the mean of the true values, Represents the predicted value, and N represents the number of samples.

Images