Tabular

• Machine learning models on tabular data are often custom-made on company-specific data by machine learning experts. Models are trained on events like customers than churn or buy together with associated information like recent customer care contacts, size of contract, last purchases, etc.
• AutoML services now support building machine learning models by just providing historic datasets, indicating what should be predicted and let the machine do the rest
• Cross-industry services now support building machine learning models by just providing historic datasets, indicating what should be predicted and let the machine do the rest
• Cluster data around norms

What to use it for

• Use AutoML to predict if a customer will churn or respond to upsell
• Use a purpose-built fraud detection model to identify fraudulent accounts and online payments
• Forecast demand for a variety of related goods with deep time-series models
• Detect anomalies within business metrics or in production data
• Segment customers into groups with similar attributes
• ...