通过模型的大规模测试预测毒性测定

Question

我目前正在创建一个模型来预测生物体污染的毒性测定。

由于没有合适的数据集，尚未尝试。我只是想问我的代码是否正确。欢迎所有批评。另外，如果我遗漏或应该包含任何内容，请告诉我。

此外，我也在考虑更多的模型，例如 RandomForestRegressor、Boosting（AdaBoost、GradientBoost）。我应该考虑这些吗？另外，当我最终获得数据时，是否应该从测试中删除任何模型？

import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

df = pd.read_csv('') # csv file with data (concentration and mortality)

# basic chart
sns.scatterplot(data = df, x = 'Concentration', y = 'Mortality') 

# basic split of training vs testing
from sklearn.model_selection import train_test_split 
X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.1, random_state=101) 

# Linear model
from sklearn.linear_model import LinearRegression 
lr_model = LinearRegression()
lr_model.fit(X_train,y_train)
lr_preds = lr_model.predict(X_test)
from sklearn.metrics import mean_absolute_error, mean_squared_error
mean_absolute_error(y_test, lr_preds)
np.sqrt(mean_absolute_error(y_test, lr_preds))
concentration_range = np.arange(0,100) # adjustable based on min/max concentration
concentration_preds = lr_model.predict(concentration_range.reshape(-1,1))
plt.figure(figsize = (12,6),dpi = 200)
sns.scatterplot(data = df, x = 'Concentration', y = 'Signal')
plt.plot(concentration_range,concentration_preds)

# Polynomical model
# Function for testing models
def run_model(model, X_train, y_train, X_test, y_test):
    # Fit model
    model.fit(X_train,y_train)

    # Get metrics
    preds = model.predict(X_test)
    rmse = np.sqrt(mean_squared_error(y_test,preds))
    mae = mean_absolute_error(y_test, preds)
    print(f'MAE: {mae}')
    print(f'RMSE: {rmse}')

    # Plot results model signal range
    concentration_range = np.arange(0,100) # again this is adjustable
    concentration_preds = model.predict(concentration_range.reshape(-1,1))

    plt.figure(figsize = (12,8), dpi = 200)
    sns.scatterplot(x = 'Concentration', y = 'Mortality', data = df, color = 'black')
    plt.plot(concentration_range, concentration_preds)

from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import PolynomialFeatures

pipe = make_pipeline(PolynomialFeatures(degree = 2),LinearRegression()) # degree is adjustable
run_model(pipe, X_train, y_train, X_test, y_test)

# K-Nearest Neighbors model
from sklearn.neighbors import KNeighborsRegressor
k_values = [1,2,3,4,5,6,7,8,9,10]
for k in k_values:

    model = KNeighborsRegressor(n_neighbors=k)
    run_model(model, X_train,y_train,X_test, y_test)

# Decision Tree model
from sklearn.tree import DecisionTreeRegressor
model = DecisionTreeRegressor()
run_model(model, X_train, y_train, X_test, y_test)

# SVR Model
from sklearn.svm import SVR # Support Vector Regression
from sklearn.model_selection import GridSearchCV
svr = SVR()
param_grid = {'C':[0.01,0.1,1,5,10,100,1000],
              'gamma':['auto','scale']}

grid = GridSearchCV(svr, param_grid)
run_model(grid, X_train,y_train,X_test, y_test)

Answer 1

您的代码看起来不错，我会考虑添加 R^2 以获得整体适合。