With the changing market landscape, understanding market dynamics has never been more important. Supply and demand will vary drastically across different markets, it is vital to forecast these ​spatial and temporal trends ​as hotels begin to plan for the rebound in travel demand. At LodgIQ, we are enabling a paradigm shift towards the usage of undiscovered and alternative data using state-of-the-art machine learning and artificial intelligence methods that will forecast recovery in real-time (Occupancy, ADR, RevPAR) for the next 12 months. 

The total scope of the COVID-19 is yet to be understood. The supply of testing has not yet met demand. The economy is experiencing a severe retraction in growth sending the unemployment rate skyrocketing to never seen before levels indicating that travel will be impacted for a prolonged period. To adapt to this COVID-19 reality, a novel forecasting framework is required to explain the consumer traveling behavior as historical trends are no longer applicable leading to the empirical loss of traditional and rigid forecasting methods. Our new forecasting framework must take into account these unforeseen patterns in order to forecast demand in both the short term and long-term.

At LodgIQ, we have continuously thought about how we can leverage Big Data and develop a forecasting framework to reflect the COVID-19 environment. Unlike the last two down turns, where data was very sparse and difficult to collect, there is a vast supply of data available and accessible to analyze and model the path forward. LodgIQ has created a causal forecasting framework​, where we monitor COVID-19 conditions in the world’s major markets, use epidemiological and macroeconomic data, and explore their effects on travel demand and simulate now-casting. By leveraging the data signals buried in the digital world and past downturns, LodgIQ has adopted a ​hybrid approach of machine learning and statistical modeling​. Boosting machines along with linear quadratic estimators are utilized to model complex nonlinear feature interactions. This new forecasting framework ensures we have an end-to-end time series forecasting framework that reduces latency, scales both horizontally (markets), and vertically (dynamic feature selection) handles model decay and supports real-time data processing.