mfn

Implementation in Python of the MFN method to measure complexity & entropy of time series. This is useful if you need features for a model such as Porfolio Optimization, clustering of time series, etc. This method is the implementation of the paper Scientific progress in information theory quantifiers. (Chaos, Solitons & Fractals, 170, 113260., Martins, A. M. F., Fernandes, L. H. S., & Nascimento, A. D. C. (2023).)

Installation

pip install mfn

or using Poetry

poetry add mfn

Usage

import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from mfn.entropy import MFN

## Generating a time series with trend and noise.
time_series = np.arange(0, 100, 1)
time_series = time_series + np.random.normal(0, 10, size=len(time_series))

value_dict = MFN(
    time_series,
    b=10,
    B=.1,
    size=100,
    dx=3
)

f, ax = plt.subplots(figsize=(6, 6))
value_df = pd.DataFrame(value_dict).reset_index()
value_df = value_df.melt(id_vars='index', value_vars=value_df.columns[1:])
sns.barplot(value_df, x='variable', y='value', errorbar="sd")
plt.title("MFN method results")
f.tight_layout()
plt.show()