Time Series Forecasting using ARIMA & Exponential Smoothing

A complete Time Series Forecasting project developed as part of my Data Science Summer Internship 2026. This project analyzes the famous Airline Passengers dataset to understand trend, seasonality, and forecasting using classical time series models.

---

Project Overview

This project demonstrates how to analyze and forecast monthly airline passenger traffic using Time Series Analysis techniques.

The project includes:

• Data preprocessing
• Time series visualization
• Trend & seasonality analysis
• Time series decomposition
• Exponential Smoothing forecasting
• ARIMA forecasting
• Model performance comparison using MAE and RMSE

---

Project Structure

TIME-SERIES-FORECASTING
│
├── data/
│   └── airline-passengers.csv
│
├── images/
│   ├── time_series.png
│   ├── decomposition.png
│   ├── exponential_smoothing_forecast.png
│   └── arima_forecast.png
│
├── time_series_forecasting.py
├── requirements.txt
├── README.md
└── .gitignore

---

Dataset

Dataset: Airline Passengers Dataset

• Monthly passenger data
• Years: 1949 – 1960
• Total Records: 144
• Missing Values: None

---

Technologies Used

• Python
• Pandas
• NumPy
• Matplotlib
• Seaborn
• Statsmodels
• Scikit-learn

---

Time Series Visualization

The original time series clearly shows an increasing trend and strong yearly seasonality.

---

Time Series Decomposition

The dataset is decomposed into:

• Trend
• Seasonal Component
• Residual (Noise)

---

Exponential Smoothing Forecast

Holt-Winters Exponential Smoothing was used to forecast future passenger counts while capturing trend and seasonality.

---

ARIMA Forecast

An ARIMA forecasting model was also implemented for comparison.

---

Model Performance

Model	MAE	RMSE
Exponential Smoothing	20.85	25.81
ARIMA	67.39	86.81

---

Key Insights

• Passenger traffic shows a strong upward trend.
• Seasonal peaks occur every year.
• Time series decomposition successfully separates trend, seasonality, and residual components.
• Exponential Smoothing produced significantly lower forecasting errors than ARIMA.
• Holt-Winters Exponential Smoothing is better suited for datasets with both trend and seasonality.

---

How to Run

Clone the repository

git clone https://github.com/zaarakhan-dev/TIME-SERIES-FORECASTING

Go to the project folder

cd TIME-SERIES-FORECASTING

Install dependencies

pip install -r requirements.txt

Run the project

python time_series_forecasting.py

---

Learning Outcomes

Through this project I learned:

• Time Series Analysis
• Trend Identification
• Seasonal Decomposition
• Exponential Smoothing
• ARIMA Modeling
• Forecast Evaluation
• MAE & RMSE Metrics
• Time Series Visualization

---

Author

Zaara Khan

B.Tech Computer Science Engineering

Data Science Summer Internship 2026