Skip to content

Forecasting Atmospheric CO2 and Forecasting Restaurant Visitors using classical (Holt-Winter's and ARIMA) and contemporary (RNNs and LSTMs) approaches

Notifications You must be signed in to change notification settings

raamav/Time-Series-Analysis

Repository files navigation

Time-Series

Forecasting Atmospheric CO2 and Forecasting Restaurant Visitors using classical (ARIMA) and contemporary (RNNs) approaches

Forecasting Atmospheric CO2

Forecasting Atmospheric CO2: For this case-study, we are utilizing the CO2 concentrations at the Mauna Lao observatory in Hawaii. The CO2 concentrations are recorded in ppm since the last 1950s (for the last 60+ years).

The goal is to forecast the monthly average CO2 concentration of 2018 and 2019. The CO2 concentrations at Sepetember 2018 and Sepetember 2019 were 405.51ppm and 408.54ppm respectively. So, for these months, we'll see if we are able to get forecasts that are reasonably accurate.

Restaurant Visitor Forecasting

Being able to predict restaurant visits accuractly is a neat application of forecasting techniques. In this case, we have the data for visits for a restaurant on a daily basis for 1+ years. From an analytic standpoint, a more nuanced version of forecasting will be attempted, one which incorporates exogenous variables.

This data is an adapted version of a recently concluded competition on Kaggle

Forecasting Alcohol Sales

The USP of the analysis is the three methods that have been used to get the predictions for the time-series, starting with Holt-Winter's (Triple Exponential Smoothing), to ARIMA (trained via Grid Search) finally culminating at forecasting using LSTMs.

Interestingly, it is the ARIMA based model that performs the best

About

Forecasting Atmospheric CO2 and Forecasting Restaurant Visitors using classical (Holt-Winter's and ARIMA) and contemporary (RNNs and LSTMs) approaches

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published