thompn/forecast

simple time series regression which takes in a csv file and uses prophet to forecast

Forecast Daily Total Volumes

This code uses the Prophet library to forecast daily total volumes using data from a file specified as an argument.

For now, this only covers daily simplistic forecasting.

Requirements

• Python 3.6 or higher
• Pandas
• Prophet
• Matplotlib
• Sklearn

Usage

First set up your virtual environment. Copy the following code into a terminal:

python -m venv venv

source venv/bin/activate

pip install -r requirements.txt

Now make sure you have a csv file in the same directory as the python file. This csv file can include any number of columns, but it must include the following columns: date (a column of dates) volume_to_predict (this can have any name, and can be specified in the arguments passed when running the python script)

Now you're ready to run, use the following command to run the forecast:

python fc.py filename.csv date_column column_to_predict

Where filename.csv is the path to the csv file containing the data to be used to make the forecast, date_column is the name of the date column in your input data, and column_to_predict is the name of the column containing the values you wish to predict.

For example if your csv file is called sales.csv and has the columns date and sales_volume then you would run as:

python fc.py sales.csv date sales_volume

Alternatively, you can run it more verbosely as follows:

python forecast.py --filename data.csv --date_column date --column_to_predict volume

Output

The code will output two files:

• forecast_output_yyyy-mm-dd.csv: a csv file containing the forecasted total volumes for each day starting from the day after the last day in the input data.
• forecast_yyyy-mm-dd_plots.pdf: a pdf file containing four plots:
  • A plot of the actual total volumes used to make the forecast.
  • A plot of the forecasted total volumes.
  • A plot showing the components of the forecast, including the trend and weekly and yearly seasonality.
  • The residuals of the fitted model

The code will also print the mean absolute error (MAE) and mean squared error (MSE) of the forecast to the terminal.