Datasets are the key building block of Predictive Insights. Each dataset is a collection of data that represents customers and prospects. This list of accounts is used for scoring and enrichment of models.

The creation of a dataset is the first step in Predictive Insights as it'll store the data required for further actions or model training.

There are two types of datasets in Predictive Insights. The first type is for scoring or enriching and another for model training in addition to scoring and enrichment. A dataset meant for scoring and enrichment is the most common use case and depends only on the existing list.

The second type of dataset is meant for model training, in addition to scoring and enrichment. Model training is the process in which a machine learning algorithm uses account information to detect patterns and relationships between attributes and customer or prospect accounts in a dataset to generate predictions for prospective accounts.

A dataset meant for Predictive Insights model training, in addition to scoring and enrichment, has additional requirements. Learn about these requirements and how to create datasets.