regressor_tester_snip.txt

def regressor_tester(reg_, X_tr, X_te, y_tr, y_te, verbose=False, display_res=False, keep_preds=False):
    import pandas as pd
    import numpy as np
    from sklearn import metrics
    
    ## Check if multiple regressors to check
    if isinstance(reg_, list):
        ## Container for multiple regressor results + counter
        count = 0
        holder = []
        
        ## Looping through each regressor
        for r in reg_:
            ## Fit/predict on train/test data
            y_hat_train, y_hat_test = fit_n_pred(r, X_tr, X_te, y_tr, show_reg=verbose)
            ## Store/get training data metrics
            tr_mse = metrics.mean_squared_error(y_tr, y_hat_train)
            tr_rmse = np.sqrt(metrics.mean_squared_error(y_tr, y_hat_train))
            tr_r2 = metrics.r2_score(y_tr, y_hat_train)
            ## Store/get testing data metrics
            te_mse = metrics.mean_squared_error(y_te, y_hat_test)
            te_rmse = np.sqrt(metrics.mean_squared_error(y_te, y_hat_test))
            te_r2 = metrics.r2_score(y_te, y_hat_test)
            ## Creating structure for df in list format
            reg_res = [['Name', 'Tr_MSE', 'Tr_RMSE', 'Tr_R2', 'Te_MSE', 'Te_RMSE', 'Te_R2'], 
                       [str(type(r)), tr_mse, tr_rmse, tr_r2, te_mse, te_rmse, te_r2]]
            ## Convert list into dataframe + append to holder
            res_df_inner = pd.DataFrame(reg_res[1:], columns=reg_res[0])
            holder.append(res_df_inner)
            count += 1
        
        ## Display counter + merge dfs into one  
        print(('---'*10), f'{count} regressors evaluated.', ('---'*10))
        res_df = pd.concat(holder, ignore_index=True)
        ## Optional Q.C.
        if display_res:
            display(res_df)
        ## Optional return of preds
        if keep_preds:
            return res_df, y_hat_train, y_hat_test
        
        return res_df
    
    else:
        ## Fit/predict on train/test data
        y_hat_train, y_hat_test = fit_n_pred(reg_, X_tr, X_te, y_tr, show_reg=verbose)
        ## Store/get training data metrics
        tr_mse = metrics.mean_squared_error(y_tr, y_hat_train)
        tr_rmse = np.sqrt(metrics.mean_squared_error(y_tr, y_hat_train))
        tr_r2 = metrics.r2_score(y_tr, y_hat_train)
        ## Store/get testing data metrics
        te_mse = metrics.mean_squared_error(y_te, y_hat_test)
        te_rmse = np.sqrt(metrics.mean_squared_error(y_te, y_hat_test))
        te_r2 = metrics.r2_score(y_te, y_hat_test)
        ## Creating structure for df in list format
        reg_res = [['Name', 'Tr_MSE', 'Tr_RMSE', 'Tr_R2', 'Te_MSE', 'Te_RMSE', 'Te_R2'], 
                   [str(type(reg_)), tr_mse, tr_rmse, tr_r2, te_mse, te_rmse, te_r2]]
        ## Convert list into dataframe
        res_df_ = pd.DataFrame(reg_res[1:], columns=reg_res[0])

        ## Optional Q.C.
        if display_res:
            display(res_df_)
        ## Optional return of preds
        if keep_preds:
            return res_df_, y_hat_train, y_hat_test
        
        return res_df_
        
## NECESSARY CODE FOR PREDICTION ##

def fit_n_pred(reg_, X_tr, X_te, y_tr, show_reg=True):
    ## Try/Except if timer object available
    try:
        timer = Timer()
        timer.start(disp_time=False)
    except:
        print('No timer for fitting.')
    
    ## Fit model to training data
    reg_.fit(X_tr, y_tr)
    
    ## Get training/test predictions
    y_hat_trn = reg_.predict(X_tr)
    y_hat_tes = reg_.predict(X_te)

    ## Try/Except if timer available
    try:
        timer.stop(disp_time=False)
    except:
        pass
    
    ## Optional Q.C.
    if show_reg:
        display(reg_)

    return y_hat_trn, y_hat_tes