jpandas.py 文件源码

def pd_gscv( pdr, method, xM, yV, alphas_log, colname = 'Predicted-RP', fname = 'sheet/rafa36795_cxcalc_prp1000.csv'):
    """
    This run grid search, perform cross-validation for plotting and save the predicted values,
    """

    print("1. Searching the best hyper-parameter by a grid method.")
    gr = jgrid.gs( method, xM, yV, alphas_log)
    print(gr.grid_scores_)
    print("Best alpha:", gr.best_params_['alpha'])

    print("2. Predicting the property using the best hyper-parameter and show a x-y plot")
    yV_pred = jgrid.cv( 'Lasso', xM, yV, alpha = gr.best_params_['alpha'], grid_std = gr_beststd(gr))

    print("3. Saving the predicted results in crossvalidation into", fname)
    pdw = pdr.copy()
    pdw[ colname] = yV_pred.tolist()
    pdw.to_csv( fname, index = False)

    print("4. Saving the best estimator as a pkl file")
    print(gr.best_estimator_)
    externals.joblib.dump(gr.best_estimator_, fname[:-3] +  "pkl")