Common Steps for using Machine Learning Model

Common Steps to Use a Machine Learning Model

1) Load the data & Split data into X & y 

import pandas   
pandas.read_csv("./data/1.csv")
X = df.drop("target", axis=1) # using all columns besides target
y = df["target"] #  predicting y using X

2) Model selection  & Split the data into training and test sets

from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(X, y,test_size=0.2)

3) Instantiate the Model  

from sklearn.ensemble import RandomForestClassifier
m= RandomForestClassifier(n_estimators=50)

4) Fit the model using function 

m.fit(X_train,y_train); 

5) Make prediction

ypreds=m.predict(X_test)

7) To Evaluate Model use  score() function on test and train data

m.score(X_test, y_test)
m.score(X_train,y_train)

8) To improve the model by changing its hyperParameters

from sklearn.model_selection import cross_val_score

import numpy as np

# Use different numbers of  n_estimators  as hyperparameter 

np.random.seed(40)
for i in range(10, 100, 5):
print(f"Trying model with {i} estimators...")
m= RandomForestClassifier(n_estimators=i).fit(X_train, y_train)
print(f"Model accruacy on test data set: {m.score(X_test, y_test)}")

9) Save trained model to file using pickle

import pickle
pickle.dump(m, open("My_Random_forest_model.pkl", "wb"))

10) Load a saved model and make a prediction on a single example
saved_model = pickle.load(open("My_Random_forest_model.pkl", "rb"))
saved_model.score(X_train,y_train)
#######################Example of  a Machine Model###########
Items_List =final_df.ITEMS.unique() ##ITEMS is the column in dataframe having the Item_names
for x in range(0,1): #x in Items_List
    y='Item_names'
    fc=get_test()
    print(fc)
    fc=fc[['ds', 'yhat', 'yhat_lower', 'yhat_upper']]
    fc['Region_NAME']=y    
    fc.to_csv('./Future_ITEMS/'+y+'.csv')#Future_ITEMS is a folder
	
	
def get_test():
  data=final_df#.loc[final_df['ITEMS']==x]
  param_grid = {'changepoint_prior_scale': [0.001, 0.01, 0.1, 0.5],'seasonality_prior_scale': [0.01, 0.1, 1.0, 10.0],}
  # Generate all combinations of parameters
  all_params = [dict(zip(param_grid.keys(), v)) for v in itertools.product(*param_grid.values())]
  rmses = []  # Store the RMSEs for each params here
    for params in all_params:
            m = Prophet(**params).fit(data)  # Fit model with given params
            df_cv = cross_validation(m, initial='180 days', period='90 days', horizon = '365 days')
            print(df_cv)
            df_p = performance_metrics(df_cv, rolling_window=1)
            print(df_p)
            rmses.append(df_p['rmse'].values[0])
# Find the best parameters
tuning_results = pandas.DataFrame(all_params)
tuning_results['rmse'] = rmses
best_params = all_params[np.argmin(rmses)]
# Initiate model with best parameters
m = Prophet(changepoint_prior_scale=best_params['changepoint_prior_scale'], 
seasonality_prior_scale=best_params['seasonality_prior_scale']).fit(data)
# make prediction
future = m.make_future_dataframe(periods=1460,freq='D') 
forecast = m.predict(future)
return forecast