from matplotlib import pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.neural_network import MLPClassifier
import pandas as pd
import numpy as np

data = pd.read_csv('sberbank_data.csv', index_col='id')
x = data[['timestamp', 'full_sq', 'floor', 'max_floor', 'build_year', 'num_room', 'material', 'kremlin_km']]

x = x.replace('NA', 0)
x.fillna(0, inplace=True)

col_date = []

for val in x['timestamp']:
    col_date.append(val.split('-', 1)[0])

x = x.drop(columns='timestamp')
x['timestamp'] = col_date

y = []
for val in data['price_doc']:
    if val < 1500000:
        y.append('low')
    elif val < 3000000:
        y.append('medium')
    elif val < 5500000:
        y.append('high')
    elif val < 10000000:
        y.append('premium')
    else:
        y.append('oligarch')

x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.01, random_state=42)

min_scores = []
med_scores = []
max_scores = []


def do_test(iters_num):
    global x_train, x_test, y_train, y_test, min_scores, med_scores, max_scores

    print("Testing iterations number "+str(iters_num)+":")
    scores = []

    for i in range(10):
        neuro = MLPClassifier(max_iter=200)
        neuro.fit(x_train, y_train)
        scr = neuro.score(x_test, y_test)
        print("res"+str(i+1)+": "+str(scr))
        scores.append(scr)

    print("Medium result: "+str(np.mean(scores)))

    min_scores.append(np.min(scores))
    med_scores.append(np.mean(scores))
    max_scores.append(np.max(scores))


def start():
    global min_scores, med_scores, max_scores

    iter_nums = [200, 400, 600, 800, 1000]

    for num in iter_nums:
        do_test(num)

    plt.figure(1, figsize=(16, 9))
    plt.plot(iter_nums, min_scores, c='r')
    plt.plot(iter_nums, med_scores, c='b')
    plt.plot(iter_nums, max_scores, c='b')
    plt.show()


start()