Introduction to Time series

Understand the basics of time series

Importing the dataset and required libraries

Exploratory Data Analysis (EDA)

White Noise detection

Random Walk detection

Stationarity test

Seasonality plot

Holt Winter Exponential Smoothing model

ARIMA model

ACF plots

Log-likelihood and AIC test

ARIMAX model

SARIMAX model

**Business Objective **

** **

A time series is simply a series of data points ordered in time. In a time-series, time is often the independent variable, and the goal is usually to make a forecast for the future.

Time series data can be helpful for many applications in day-to-day activities like:

- Tracking daily, hourly, or weekly weather data
- Monitoring changes in application performance
- Medical devices to visualize vitals in real-time

Auto-Regressive Integrated Moving Average (ARIMA) model is one of the more popular and widely used statistical methods for time-series forecasting. ARIMA is an acronym that stands for Auto-Regressive Integrated Moving Average. It is a class of statistical algorithms that captures the standard temporal dependencies unique to time-series data.

The model is used to understand past data or predict future data in a series. It’s used when a metric is recorded in regular intervals, from fractions of a second to daily, weekly or monthly periods.

ARIMAX (Auto-Regressive Integrated Moving Average Exogenous) is an extension of the ARIMA model, and similarly, SARIMAX (Seasonal Auto-Regressive Integrated Moving Average with Exogenous factors) is also an updated version of the ARIMA model. We will see how to implement these two models as well.

We have already covered the concepts of Autoregression modeling and Moving Average Smoothing techniques.

In this project, we will be implementing the ARIMA model on the given dataset.

**Data Description **

The dataset is “Call centers” data. This data is at the month level wherein the calls are segregated at the domain level as the call center operates for various domains. There are also external regressors like no of channels and no of phone lines which essentially indicate the traffic prediction of the inhouse analyst and the resources available.

The total number of rows are 132 and the number of columns are 8:

- Month, healthcare, telecom, banking, technology, insurance, no of phonelines and no of channels.

**Aim**

** **

This project aims to build an ARIMA model on the given dataset.

**Tech stack **

- Language - Python
- Libraries - pandas, numpy, matplotlib, seaborn, statsmodels, scipy

** **

**Approach **

- Import the required libraries and read the dataset
- Perform descriptive analysis
- Exploratory Data Analysis (EDA) -

- Data Visualization

- Check for white noise
- Check for Random Walk
- Perform Stationarity tests

- Augmented Dickey-Fuller test
- KPSS test

- Seasonal decomposition plot
- Holt Winter Exponential Smoothing

- Create and fit the model
- Make predictions on the model
- Plot the results

- ARIMA model

- Create models with varying lag values
- Compare these models using log-likelihood and AIC values
- Check with the LLR test
- ACF Plots of residuals

- ARIMAX model

- Create a model
- ACF plots of residuals

- SARIMAX model

- Create a model
- ACF plots of residuals

Time series introduction

06m

Dataset overview

06m

Check for the White Noise in the data

07m

Check for Random Walk in the data

07m

Check for Stationarity in the data

06m

Check for Seasonality in the data

07m

Holt Winter Exponential Smoothing method

06m

Overview and limitations of ARMA model

06m

Introduction to ARIMA model

07m

Working of ARIMA model

08m

Implementation of ARIMA model

07m

Evaluating the ARIMA model

07m

ARIMAX and SARIMAX models

08m

Modular code overview

04m