Explore Data

After you've added a dataset and selected a target, as described here, it's time to explore the data set with Kortical's ML Data Prep feature. This process produces a report containing detailed data insights, as well as a set of recommended actions to apply to transform the data in preparation for model creation.

The following sections explain how the process works and describes in detail the components of the exploration report.

What is ML Data Prep?

ML Data Prep is a machine learning driven automated EDA (Exploratory Data Analysis) tool, which aims to save a lot of the initial grunt work a data scientist has to perform when first analysing a dataset.

Under the surface, a huge array of different machine learning models work together to guide users to the most important insights, provide an indication of model performance and even suggests any features which can be dropped to improve performance.

ML Data Prep consists of four main components:

• Data Completeness - which highlights data quality issues and missing values. It will also detect columns with no information which can be safely discarded.
• Column Insights - which ranks features by their importance to the target and allows many different views on the fundamental structure of each input column. It also performs leak detection and Feature Selection, both of which can have a big impact on the performance of downstream models.
• Target Insights - which gives indicative model performance and helpful suggestions of which evaluation metric may be most appropriate, among other things.
• Actions Summary - this allows the user to transform their input dataset based on the suggested given by the above components.