In recent years, big data has been widely used to understand emerging trends in several markets. With this paper, we consider the use of real estate websites data and machine learning in the problem of knowing the current fair value of banking collateral in Morocco. In our study, we are going to utilize the online data, containing housing sales advertisements, extracted from three websites: (Avito.ma, Mubawab.ma and Sarouty.ma), popular online portals for real estate services in Morocco, in order to develop forecasting models of apartment prices using artificial intelligence for the city of Casablanca. Each record of the database contains information on the listed housing unit (asking price, district, floor, area, number of rooms, balcony, terrace…), on the building (elevator, garage, garden, etc.), on the ad (publication date, apartment sector) and a short description. After collecting the database and the deduplication process, we make additional controls on the dataset to address potential incoherence errors in the data. After validating the clean dataset against official statistical sources, we construct forecasting models for real estate sale prices, based on artificial intelligence and statistical methodologies.