A simple yet challenging project, to forecast the volume of air passengers, based on monthly totals of a US airline passengers from 1949 to 1960. Can you overcome the obstacles & Forecast the future occupancies of the Airlines?
This data frame contains the following columns:
- Month : The month of observation
- #Passengers : Total Passengers travelled in that particular month
Source:
This dataset is taken from an inbuilt dataset of R called AirPassengers.
Kaggle -
https://www.kaggle.com/yersever/500-person-gender-height-weight-bodymassindex
- Understand the Dataset & cleanup (if required).
- Perform the necessary checks like stationarity & DF on the Dataset.
- Build a forcasting model to predict the future volumne of the air passengers.
- Visualize the time series - Check for trend, seasonality, or random patterns.
- Stationarize the series using decomposition or differencing techniques.
- Plot ACF/PACF and find (p,d,q) parameters.
- Building the forecasting model - can be AR, MA, ARMA or ARIMA.
- Making Predictions using the Forecasting Model
1. Time-Series Data
2. Stationarity Check
3. Decomposing using moving average
4. Stationarity Check for Decomposed data
5. Auto Correlation Function Plot
6. Partial Auto Correlation Function Plot
7. ARIMA Model Forecast
- The Air-Passengers Time-Series Dataset was quiet small, with just 144 samples.
- It was clear from the visuals that the time-series dataset had an upward trend & some seasonality.
- The same was confirmed with help of visual (rolling mean & std) & statistical (Dicky-Fuller Test) stionarity checks.
- The time-series was subject to Decomposition in order to stationarize the outputs.
- Futher ACF & PACF curves were plotted to extract the values of p & q, as it is required for the ARIMA Model.
- The Forecasting Model was then built with the time-series data, by feeding the optimal p,q,d values.
- Finally, the model was used to forecast the time-series of the air-passengers, into the future.