Statistical and ML models for time series analysis combined with data cleaning, data visualization and statistical inference
Jupyter Notebook 93.11%Python 0.52%TeX 6.37%
timeseriesanalysis's Introduction
Time Series Analysis
Projects:
Power Demand Foreast of South AustraliaRepresentative Project Reducing Power Supply Costs in South Australia using Statistical time series Modelling and ML Methods. (QBUS3830)
AWS Forecast Golf with Weather Cleaning and Stationary Time series data in AWS sageMaker, and then Modeling, Deploying, and Forecasting using Sagemaker DeepAR+ and Amazon Forecast console. (Own Project)
Sales Forecasting Forecast six weeks daily sales for several stores by developing a univariate forecasting model. (QBUS2820)
Amazon Forcast A managed time series forecasting service that uses AWS machine learning technology.
Study Notes (ECMT 2160)
Basic Knowledge for time series regression.
Jupyter NoteBook Contents
Data Analysis and Visualisation
DataCleaning Proper way to remove outliers for time series data
EDA and Feature Engineering Time decomposition and Seasonality Plot. Stationary check and transformation by ACF and PACF plots.
Machine Learning
NN models neural network with hyperparameter optimisation process
DeepAR+ sagemaker Amazon built-in Algo: a supervised learning algorithm for forecasting scalar (one-dimensional) time series using recurrent neural networks (RNNs) combined with classical foprecast methods such as ARIMA and Expentional smoothing.
Bootstrap CI
Bootstrapping the MAPE and MAE of neural network resuduals\
Statistical Modelling
Random Walk, SARIMA and ES models Rolling Window and Fixed Window forecast, Residual diagnostics based on ACF and PACF plots and Residual distribution (histogram and QQ plot), Point forecast measure by PAME
AR models Lag Selection by AIC with rolling window forecast.