brockmanmatt

!pip install gdelt
import gdelt
gd = gdelt.gdelt(version=1)

import os, datetime

os.makedirs("data",exist_ok=True)

#starting 60 days before Oct7
cur_date = datetime.datetime(2019,10,7)-datetime.timedelta(days=60)

brockmanmatt

mySources = ["nytimes.com", "washingtonpost.com", "foxnews.com", "cnn.com"]

df = pd.DataFrame()
k = os.listdir("data")
for i in k:
  print(i)
  if i.endswith(".pkl"):
    tmp = pd.read_pickle("data/"+i)
    tmp = tmp[tmp["SOURCES"].apply(lambda x: x in mySources)]
    df = pd.concat([df, tmp])

brockmanmatt

def get_supervised_df(df, issue, maxlag=1):
  
  original = df.copy()
  
  #for each day of lag, we add a new column to the dataframe with the previous values
  for i in range(maxlag):
    print(i)
    original= pd.concat([original, df.shift(i+1).add_suffix("_(t-%s)"%(i+2))], axis=1).dropna()

  #save original issue to add back in

brockmanmatt

from fastai.tabular import *

def genNN(myDF, issue):  
  reframed = get_supervised_df(myDF, issue, 7)


  days = []
  myOrder = reframed.columns.to_list()

  for i in range(7):

brockmanmatt

h_predictions, *_ = learn.get_preds(DatasetType.Test)
#h_predictions = h_predictions.numpy()

v_predictions, *_ = learn.get_preds(DatasetType.Valid)
#h_predictions = h_predictions.numpy()

predictions, *_ = learn.get_preds(DatasetType.Train)
#predictions = predictions.numpy()

issue = "foxnews.com_russia"

brockmanmatt

from sklearn.metrics import mean_squared_error
from math import sqrt

testing = results.copy()

print("RMSE training  using mean: {}".format(sqrt(mean_squared_error(testing.actual, testing.m))))
print("RMSE training using model: {}".format(sqrt(mean_squared_error(testing.actual, testing.predicted))))

testing = results[-len(h_predictions)-1:].copy()

brockmanmatt

fig, axs = plt.subplots(4, 5, sharex=True)

fig.set_size_inches(16,12)

x = y = 0

for issue in time_series:
  train_l = 55

  s_model = SARIMAX(endog = time_series[[issue]][:train_l][1:],

brockmanmatt

myCountries = ["dprk", "ukraine", "russia", "iran", "china"]

fig, axs = plt.subplots(1,5, sharey=True, gridspec_kw={'wspace': 0})

fig.set_facecolor("white")
fig.set_size_inches(24,6)
idx = 0

for country in myCountries:
  tmp = time_series[[x for x in time_series.columns if (x.find(country) > -1) ]]

brockmanmatt

fig, axs = plt.subplots(1, 5, sharey=True, gridspec_kw={'wspace': 0})

fig.set_size_inches(16,6)

x = y = 0

for issue in myCountries:
  train_l = len(time_series)-5

brockmanmatt

#create a DF to hold errors
search_df = pd.DataFrame()

#set up the grid (1 row x 5 columns)

fig, axs = plt.subplots(1, 5, sharey=True, gridspec_kw={'wspace': 0})

fig.set_size_inches(16,6)

x = y = 0