successor

Uses pre-trained tensorflow models to predict the next k entries in a sequence

Install

pip install successor

You may get better performance by first installing tensorflow following the instructions and perhaps reading this thread.

Use

See basic_use

# 1. Import a skater
from successor.skaters.scalarskaters.scalartsaskaters import successor_tsa_aggressive_d0_ensemble as f

# 2. Univariate data
import numpy as np
y = list(np.cumsum(np.random.randn(1000)))

# 3. Initialize state to empty dict
s = {}

# 4. Give it some data (observations) one at a time, each time passing it back the state s
for yi in y:
    x, x_std, s = f(y=yi,s=s,k=1)

Skaters follow the convention established by the timemachines library and you are encouraged to read the description of the "skater" signature if anything is confusing.

Benchmarking

See Elo ratings