[PYTHON] TIME SERIES FORECASTING

Introduction

• This project was created to forecast the amount of CO2.
• This data was measured from 1958 to 2001.
• Data here: https://github.com/TranThanhTuan2509/Time-Series-Forecasting/blob/main/co2.csv
• Time series forecasting could be used for Stock Forecasting, Financial Forecasting, Energy Consumption Forecasting, Weather Forecasting, etc..

The strategy to predict

• Recursive method
• Direct method

Preprocessing

• This dataset currently comprises only 2 columns, so you need to add more columns behind CO2 columns to be suitable with the 2 methods above. The number of columns to be added depends on your specific requirements.
• All Time-information columns have to change the datatype to "Datetime".
• Here is the data virtualization before handling missing data:
• As you can see, the data follows a pattern, so you do not have to handle the missing data by using simple imputation or filling it with a specific value. The only way to handle this is by using the interpolation method
• Here is the data virtualization after using the interpolation method:
• In time series forecasting, the train-test-split method should not be used during training because time series data has continuous properties, whereas train-test-split randomizes the data

Requirement

• Pandas
• sk-learn