price_prediction.py

import tweepy
import praw
import coindesk
import dict_list
import pandas as pd
import word2vec as w2v

#Step 1 - Retrieve Tweets past 30 days
twitter_config = json.load(TWEEPY_CONFIG_FILE)
auth = tweepy.OAuthHandler(**twitter_config)
api = tweepy.API(auth)
public_tweets = api.search('bitcoin')

#Step 2 - Retrieve reddit posts past 30 days
reddit_config = json.load(open(PRAW_CONFIG_FILE))
reddit = praw.Reddit(**reddit_config)
top_posts = reddit.subreddit('bitcoin').top()

#Step 3 - Retrieve bitcoin data past 30 days
price_list = coindesk.prices('1-1-2017', '30-1-2017')
for i in enumerate(price_list):
    if price_list[i] < price_list[i+1]  # did the price increase or decrease?
    	price_list[i] = 0
    else
    	price_list[i] = 1 

#Step 4 Clean data
input_data = public_tweets + top_posts
final_data
for i in input_data
	for x in dict_list
		if input_data[i] = dict_list[x]
			final_data = input_data[i]

#Step 5 Vectorize Data
vectorized_data = w2v(final_data)

#Step 6 Build LSTM Network model
model = Sequential()
model.add(LSTM(100, input_shape=(trainX.shape[1], trainX.shape[2])))
model.add(Dense(1))
model.compile(loss='mae', optimizer='adam')
trained_model = model.fit(vectorized_data, price_list, epochs=300, batch_size=100)