#import streamlit as st
import pandas as pd
import numpy as np
import csv
import json
import matplotlib.pyplot as plt
import ast
#import pickle
import sklearn
from sklearn import linear_model


df = pd.read_csv('emily_election.csv')


#loaded_model = pickle.load(open(filename, 'rb'))


df['runtime'] = df['cumulative_ad_runtime'].apply(lambda s: int(s.split('days')[0]))
df['impressions'] = df['cumulative_impressions_by_region'].apply(lambda d: ast.literal_eval(d))
df['impressions'] = df['impressions'].apply(lambda d: np.array(list(d.values())).sum())

#feature 3 (for later)
df['audience_size'] = df['cumulative_est_audience'].apply(lambda d: ast.literal_eval(d))
df['audience_size'] = df['audience_size'].apply(lambda d: np.array(list(d.values())).sum())

#data = df[['runtime', 'spend', 'impressions']]
data = df[['runtime', 'spend', 'audience_size','impressions']]

msk = np.random.rand(len(data)) < 0.8
train = data[msk]
test = data[~msk]


#new_train = train[train['impressions'] < 1000000]
train['spend'] = pd.to_numeric(train['spend'],errors='coerce')


new_train = train[(train['spend'] > 250)]
new_train = new_train[new_train['runtime']>4]
new_train.shape


#this model predicts impressions given the runtime and the spend

regr = linear_model.LinearRegression()
new_train['log_runtime'] = np.log(new_train['runtime'])
new_train['log_spend'] = np.log(new_train['spend'])
new_train['log_impressions'] = np.log(new_train['impressions'])
new_train.replace([np.inf, -np.inf], np.nan, inplace=True)
new_train.dropna(inplace=True)
x = np.asanyarray(new_train[['log_runtime', 'log_spend']])
y = np.asanyarray(new_train[['log_impressions']])
print(x)
regr.fit(x, y)
#y_pred= regr.predict(new_train[['log_runtime', 'log_spend']])


# # The coefficients
#print(regr.coef_)
#print('R-squared score: %.2f' % regr.score(x, y))
#print('Standard Deviation: %.2f' % np.sqrt(sum((y - y_pred)**2) / (len(y) - 2)))