I've been struggling to getting a feature importance of my data, which has 22 indices' variations during certain periods, and total count of typhoons during same period. And this dataframe's dimension is like (30, 22).
SF 10.7 SOI N12 N3 N34 N4 BEST lv TNA TSA WHWP ... PWR TPSE ATSE AMOL AMM NTA CAR QBO GIAM counts
Unnamed: 0
1979 1803.0 1.1 0.16 -0.15 -0.29 -0.24 -0.08 0.42 0.10 -0.81 ... -0.200 0.517 0.514 -0.051 1.96 0.11 -0.20 -19.60 0.68 2
1980 1932.0 -0.0 -0.36 0.08 0.21 0.16 0.64 0.47 -0.08 1.21 ... -0.163 0.781 0.645 0.100 3.64 0.07 0.10 7.33 0.45 3
1981 1569.0 2.0 -0.63 -0.43 -0.46 -0.59 -0.47 0.18 -0.22 -0.80 ... -0.041 -0.190 0.220 -0.066 2.19 -0.07 -0.08 -3.49 0.20 0
1982 1719.0 -1.7 0.13 0.54 0.53 0.45 1.74 -0.15 -0.45 -0.14 ... -0.273 1.468 -0.734 -0.188 -0.30 -0.24 -0.19 -15.94 1.14 3
1983 1386.0 0.1 4.24 1.52 0.54 -0.00 0.82 0.35 -0.45 4.82 ... 0.043 2.425 -0.034 -0.058 1.25 0.14 -0.04 3.12 1.38 3
This is the data i have, sorry for looking not tidy though.
I'd like to get a feature importance of this dataset. So I used a libray Shap but I do not know why I cannot get a exact result from a figure.
This is my codes.
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
# uplaod dataset
X = pd.read_csv('june_index.csv')
X.set_index('Unnamed: 0', inplace=True)
raw_data = pd.read_csv('jul_aug_TC_sum.csv')
y = raw_data.iloc[30:60, -1]
ind = X.index
y = pd.DataFrame(y.values, index=ind, columns=['counts'])
data = pd.concat([X,y], axis=1)
------------------------------------------------------------
feature_columns = list(data.columns.difference(['counts']))
XX = data[feature_columns]
yy = data['counts']
X_train, X_test, y_train, y_test = train_test_split(XX, yy, test_size = 0.3, random_state = 42)
print(X_train.shape, X_test.shape, y_train.shape, y_test.shape)
------------------------------------------------------------
import lightgbm as lgb
from math import sqrt
from sklearn.metrics import mean_squared_error
lgb_dtrain = lgb.Dataset(data = X_train, label = y_train) # 학습 데이터를 LightGBM 모델에 맞게 변환
lgb_param = {'max_depth' : 10,
'learning_rate' : 0.01,
'n_estimators' : 1000,
'objective' : 'regression'}
lgb_model = lgb.train(params = lgb_param, train_set = lgb_dtrain) # 학습 진행
lgb_model_predict = lgb_model.predict(X_test) # 평가 데이터 예측
print('RMSE : {}' .format(sqrt(mean_squared_error(lgb_model_predict, y_test))))
----------------------------------------------------
explainer = shap.TreeExplainer(lgb_model) <------ THIS IS THE PART OF PROBLEM
shap_values = explainer.shap_values(X_test)
shap.initjs()
shap.force_plot(explainer.expected_value, shap_values[1, :], X_test.iloc[1, :])
shap.force_plot(explainer.expected_value, shap_values, X_test)
please help me...