Forecasting the Space Utilization Trend in Corporate Offices

Abstract

The research is mainly focused on forecasting office space utilization trends in the organization using information such as office space count, space occupancy count, holidays, leaves. Space occupancy data is collected using PIR sensors. Descriptive analytics is done using creative visualizations, and model building is done using univariate and multivariate time series methods. Descriptive analytics explains that there is a positive autocorrelation in the data with no outliers and randomness. There exists a pattern of space occupancy for different office locations at different times of the day. Univariate time series models are suitable for forecasting space occupancy for single office locations, whereas multivariate time series model VAR is suitable when considering multiple office locations of a client or multiple office locations of different clients at the same time. Empirical research has exhibited that out of tested models, SARIMAX has shown better performance on multiple test datasets.