In data science, poor feature engineering is one of the leading culprits behind underperforming predictive models. Predictive modeling success hinges on selecting the features that are most likely to affect the desired outcome. Mobilewalla offers solutions that effectively allow data scientists to outsource aspects of feature discovery.
Through years of refinement using a wide variety of modeling use cases, we have identified nine different feature categories, predictive of a wide range of business outcomes. These features are easy to use across a wide range of business, marketing and data science applications. All of the features are mapped to the IFA (identifier for advertisers) which is the device ID or MAlD (mobile advertiser ID).
Along with the nine features categories, Mobilewalla can also generate custom features based on your specific business requirements. Talk to one of our data experts now, to learn more.
Features related to the apps being used on the device such as most seen app category, number of distinct apps used by the device, number of premium apps used by the device,etc.
Features based on device carrier like signal distribution by telco type, last seen cellular carrier and most seen WiFi carrier.
Device profiling based on usage time like night riders, commuters, early risers, weekday, and weekend engagements.
Features like area of mobility, average distance travelled in a day, total number of distinct locations visited by device, home location, work location, commute distance etc.
Features focused on related devices and device attributes for users at that same location.
Features like common day, common evening and most-seen location of devices.
Features based on POI engagement scores of devices.
Features focused on hours, days and period when the devices are active or engaged.