danilbaibak / predict_fluctuation_currency_quote

Predict fluctuations in currency quote using Prophet

Predict fluctuations in currency quote using Prophet

Overview

Let's try build model with help of Prophet to prеditi fluctuations in currency quotes. For current example I took currency pair UAH/USD for date range 2012-2017. The forecast looks quite realistic, the forecast components give some ideas about seasonality. The time series forecast can be used as a feature for another machine learning models.

Prophet works pretty well from the box. You can easily build a model for forecasting time series. Also Prophet several the hyperparameters, that can help you improve your model.

There are build in tools for visualization of the prediction. You can visulaizate the forecast and the forecast components. The visualization of the forecast components can help make interesting conclusions about seasonality.

Dependencies

• Python 3.5
• Prophet
• scikit-learn
• plotly

Installation

For installation please use conda. Just run:

make init

After the installation was done successfully, activate your enviroment:

source activate fbprophet

For using conda and environments, please read full documentation of conda.

Usage

After activate source, run jupyter notebook currency_quote_uah_usd.ipynb. The currency_quote_uah_usd.ipynb is precompiled, so you can have a look without instalation.