/Time-Series-Data-Analysis

This is a jupyter notebook using Time Series Method (Triple Exponential Smoothing) for forecasting total bandwidth of a company providing entertainment platform for music, video, live stream, chat, etc. As the dataset is only train set, so I split train set to train-tuning, val-tuning and test set.

Primary LanguageJupyter Notebook

Time-Series-Method

bandwidth_total and max_user

This is a jupyter notebook using Time Series Method (Triple Exponential Smoothing) for forecasting total bandwidth of a company providing entertainment platform for music, video, live stream, chat, etc. As the dataset is only train set, so I split train set to train-tuning, val-tuning and test set in notebook.

Bandwidth_total with SMA100 and SMA100 upper

Bandwitdh_total with SMA100 and SMA100 upper

Bandwidth_total with SMA100 and SMA100 upper after removing outliers

Bandwitdh_total with SMA100 and SMA100 upper after removing outliers

Testing for stationarity and seasonality

testing seasonality and stationarity

Calculating baseline by using Holt model

Baseline for model by Holt model

Final Triple Exponential smoothming after tunning

Triple Exponential smoothing after tunning

Red line represents model fitting train set. Blue line represents model tunning and fitting valid set. Green line represents model predicting test set.