Multioutput prediction of global cropland yields

This is a programme that predict global lnRR of cropland based on the observations.

• Agminated prediction task for globle crop prediction
• Reliable machine learning model for agricultural and research
• Outputs lives at MultiOutput_Regression/output

Tutorial

Installation

• python == 3.9.2

• sklearn == 1.2.2

• torch == 2.0.1

• pandas == 2.0.1

• numpy == 1.23.5

• scipy == 1.11.2

   pip install scikit-learn==1.2.2 torch==2.0.1 pandas==2.0.1 numpy==1.23.5 scipy==1.11.2
  

Script for replicating

bash Prediction.sh 

Read data

Read raw data from csv file.

df_raw = pd.read_csv(os.path.join(root_path, datd_path))
cols = list(df_raw.colunms);
cols = [i for i in cols if i not in target and i not in redudent]
col_y = target

df_x = df_raw[cols]
df_y = df_raw[col_y]

Build multiclass regression model

estimators = args_info['n_estimator']
random_state = args_info['random_state']
depth = args_info['max_depth']

rf = RandomForestRegressor(n_estimators=estimators, random_state=random_state, max_depth=depth)
regression = MultiOutputRegressor(rf)

Define train

train_x, test_x, train_y, test_y = train_test_split(df_x, df_y, test_size=0.2)
regression.fit(train_x, train_y)

r2_train = regression.score(train_x, train_y)
r2_test = regression.score(test_x, test_y)

print("|| Train R2: {0:.7F}    Test R2: {1:.7F}").format(r2_train, r2_test)

Do predict

Ypredicted = regression.predict(Predict_x)
preds = np.array(Ypredicted)
pred_lnRR = preds[:, 0].reshape(360, -1)
pred_substence = preds[:, 1].reshape(360, -1)

pred_lnRR = pred_lnRR * landcover
pred_substence = pred_substence * landcover

Program running

python main.py --model='Adaboost' --root_path='./raw_Data' --data_path='SOCALL.csv' --random_state=100 --traget=['lnRR', 'SOC'] --do_predict=True

Model performance