Scikit-learn has very low accuracy on LOGISTIC REGRESSION, Random Forest, SVM but has high accuracy on linear regression [closed]

Question

This is my dataset

I transformed the columns with string type to float by doing this

df2['Sex'] = df['Sex'].astype('category')
df2['Housing'] = df['Housing'].astype('category')
df2['Saving accounts'] = df['Saving accounts'].astype('category')
df2['Checking account'] = df['Checking account'].astype('category')
df2['Purpose'] = df['Purpose'].astype('category')

To train the model:

train, test = train_test_split(df2, test_size=0.2)
Y_train = pd.DataFrame()
Y_test = pd.DataFrame()
Y_train["score"] = train["score"]
Y_test["score"] = test["score"]
X_train = train.drop('score', 1)
X_test = test.drop('score', 1)


lr = LogisticRegression(penalty='l1', C=0.9, solver='liblinear', n_jobs=-1)
lr.fit(X_train, Y_train)
Y_pred = lr.predict(X_test)

My accuracy with LOGISTIC REGRESSION, RandomForest or SVM is very low

from sklearn.metrics import accuracy_score
accuracy_score(Y_test,Y_pred)

0.05

Answer 1

Your problem is regression, but you tried classification models (Logistic regression, SVM and RandomForest). You should try RandomForestRegressor, SVR (as opposed to SVC) etc.