predict.py

import numpy as np
import pandas as pd
import pickle as pkl
from sklearn.preprocessing import LabelEncoder

def allot_crops(dist_name, season, area):
    """
    Arguments : Name of the district, Season (Cropping season)
                and Farm area (input provided by user)
    Dataframe is created for all crops in original list for the given
    input, production is calculated by the model and 5 crops are output
    having production closest to predicted production
    Return : List of top 5 crops alloted
    """
    
    data = np.array([[season, area]])
    df = pd.read_csv('data/apy.csv')
    
    crops = df[df['District_Name']==dist_name]['Crop'].unique()
    data = pd.DataFrame(np.repeat(data, [len(crops)], axis=0), columns=['Season','Area'])
    data.insert(1, "Crop", crops)
    
    le1 = LabelEncoder()
    le1.classes_ = np.load('data/season.npy', allow_pickle=True)
    data['Season'] = le1.transform(data['Season'])
    
    le2 = LabelEncoder()
    le2.classes_ = np.load('data/crop.npy', allow_pickle=True)
    data['Crop'] = le2.transform(data['Crop'])
    
    file = open('models/random_forest_1.sav', 'rb')
    model = pkl.load(file)
    
    pred = model.predict(data)
    
    indices = np.argsort(pred)[::-1][:5]
    
    crops_alloted = le2.inverse_transform(data.iloc[indices]['Crop'])
    file.close()
    
    return crops_alloted