import pandas as pd
import matplotlib.pyplot as plt
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_squared_error
import numpy as np


# Read column names
col_names = []
with open('names') as f:
    col_names = f.read().strip().split('\t')

# Read data
dev = pd.read_table('dev-0/in.tsv', error_bad_lines=False,
                    header=None, names=col_names[1:])
test = pd.read_table('test-A/in.tsv', error_bad_lines=False,
                     header=None,  names=col_names[1:])
train = pd.read_table('train/train.tsv', error_bad_lines=False,
                      header=None,  names=col_names)
dev_expected = pd.read_table('dev-0/expected.tsv', error_bad_lines=False,
                             header=None)

# Create dummies for columns
for c in train.select_dtypes(include=object).columns.values:
    train[c] = train[c].astype("category").cat.codes
for c in dev.select_dtypes(include=object).columns.values:
    dev[c] = dev[c].astype("category").cat.codes
for c in test.select_dtypes(include=object).columns.values:
    test[c] = test[c].astype("category").cat.codes

# Sprawdzanie ile jest odstających wartości dla price
# fig, ax = plt.subplots(1, 2)
# fig.set_figheight(15)
# fig.set_figwidth(20)
# ax[0].boxplot(train['price'])
# ax[0].set_title('price')
# ax[1].boxplot(train['mileage'])
# ax[1].set_title('mileage')
# plt.show()

# Usunięcie odstających wartości
priceMin = 0
for price in train['price']:
    if price < 1000:
        priceMin += 1
# print("Price min cut: " + str(priceMin))

priceMax = 0
for price in train['price']:
    if price > 1000000:
        priceMin += 1
# print("Price max cut: " + str(priceMax))

mileageMin = 0
for m in train['mileage']:
    if m < 100:
        mileageMin += 1
# print("Mileage min cut: " + str(mileageMin))

train = train.loc[(train['price'] > 1000)]
train = train.loc[(train['mileage'] > 100)]

# Split train set to X and Y
X_train = train.loc[:, train.columns != 'price']
Y_train = train['price']

# Create Linear regresion model
clf = LinearRegression().fit(X_train, Y_train)

# # Predict
dev_p = clf.predict(dev)
test_p = clf.predict(test)

# # Accuracy
score = mean_squared_error(dev_p, dev_expected, squared=False)
print("RMSE: " + str(score))

# # Save to files
dev_p.tofile('./dev-0/out.tsv', sep='\n')
test_p.tofile('./test-A/out.tsv', sep='\n')