Building models using multimodal datasets in DataRobot
Having fully set up our ZIP file with the multimodal dataset, we proceed into initiating the project within DataRobot. The data ingestion using the drag and drop method is like the earlier project, except in this case we upload the ZIP file. Following the upload of the ZIP file, the price is selected as the target variable. DataRobot automatically detects the text, image, and geospatial fields (see Figure 11.2). The geometry feature is a location-based feature made up of the latitude and longitude variables in the original dataset. Apart from latitude and longitude coordinates, location features can be formed from other native geospatial formats, such as Esri shapefiles, GeoJSON, and PostGIS databases. These can be uploaded using drag and drop, AI Catalog, or URL methods:
Figure 11.4 – Feature Name list
- The location-based visual representation of the listing price can be viewed by selecting the Price option in the Feature...
