mirror of
https://github.com/kennethreitz/coinbin.org.git
synced 2026-06-05 23:10:17 +00:00
101 lines
3.0 KiB
Python
101 lines
3.0 KiB
Python
import records
|
||
import os
|
||
|
||
import maya
|
||
import numpy as np
|
||
import pandas as pd
|
||
import requests
|
||
from fbprophet import Prophet
|
||
|
||
from scraper import Coin
|
||
|
||
|
||
class MWT(object):
|
||
"""Memoize With Timeout"""
|
||
_caches = {}
|
||
_timeouts = {}
|
||
|
||
def __init__(self, timeout=2):
|
||
self.timeout = timeout
|
||
|
||
def collect(self):
|
||
"""Clear cache of results which have timed out"""
|
||
for func in self._caches:
|
||
cache = {}
|
||
for key in self._caches[func]:
|
||
if (time.time() - self._caches[func][key][1]) < self._timeouts[func]:
|
||
cache[key] = self._caches[func][key]
|
||
self._caches[func] = cache
|
||
|
||
def __call__(self, f):
|
||
self.cache = self._caches[f] = {}
|
||
self._timeouts[f] = self.timeout
|
||
|
||
def func(*args, **kwargs):
|
||
kw = sorted(kwargs.items())
|
||
key = (args, tuple(kw))
|
||
try:
|
||
v = self.cache[key]
|
||
if (time.time() - v[1]) > self.timeout:
|
||
raise KeyError
|
||
except KeyError:
|
||
v = self.cache[key] = f(*args, **kwargs), time.time()
|
||
return v[0]
|
||
func.func_name = f.__name__
|
||
|
||
return func
|
||
|
||
|
||
def get_predictions(coin):
|
||
|
||
c = Coin(coin.lower())
|
||
|
||
q = "SELECT date as ds, value as y from api_coin WHERE name=:coin"
|
||
usd = requests.get('https://coinbin.org/btc').json()['coin']['usd']
|
||
|
||
pro_db = records.Database(os.environ['HEROKU_POSTGRESQL_TEAL_URL'])
|
||
rows = pro_db.query(q, coin=c.name)
|
||
|
||
df = rows.export('df')
|
||
|
||
df['y_orig'] = df['y'] # to save a copy of the original data..you'll see why shortly.
|
||
|
||
# log-transform y
|
||
df['y'] = np.log(df['y'])
|
||
|
||
model = Prophet(weekly_seasonality=True, yearly_seasonality=True)
|
||
model.fit(df)
|
||
|
||
future_data = model.make_future_dataframe(periods=6, freq='h')
|
||
forecast_data = model.predict(future_data)
|
||
print(forecast_data[['ds', 'yhat', 'yhat_lower', 'yhat_upper']].tail())
|
||
|
||
forecast_data_orig = forecast_data # make sure we save the original forecast data
|
||
forecast_data_orig['yhat'] = np.exp(forecast_data_orig['yhat'])
|
||
forecast_data_orig['yhat_lower'] = np.exp(forecast_data_orig['yhat_lower'])
|
||
forecast_data_orig['yhat_upper'] = np.exp(forecast_data_orig['yhat_upper'])
|
||
|
||
df['y_log'] = df['y'] #copy the log-transformed data to another column
|
||
df['y'] = df['y_orig'] #copy the original data to 'y'
|
||
|
||
# print(forecast_data_orig)
|
||
d = forecast_data_orig['yhat'].to_dict()
|
||
predictions = []
|
||
for i, k in enumerate(list(d.keys())[-6:]):
|
||
w = maya.when(f'{61*i} minutes from now')
|
||
predictions.append({
|
||
'when': w.slang_time(),
|
||
'timestamp': w.iso8601(),
|
||
'usd': d[k],
|
||
})
|
||
return predictions
|
||
|
||
|
||
# results = list(zip(*[forecast_data_orig[c].values.tolist() for c in forecast_data_orig]))[-6:]
|
||
# for result in results:
|
||
# dt = maya.MayaDT(result[0])
|
||
# _min = result[7]
|
||
# _max = result[8]
|
||
# _prediction = result[16]
|
||
# print(dt.slang_time(), _min, _max, _prediction)
|