How do we make decisions? Do we know the right questions to ask? In the complex environment we live in, experience and intuition are still valuable but no longer sufficient. Decisions that work for one business do not guarantee success for another; nor will a good outcome yesterday always translate into a good one tomorrow. What if your organisation could approach decision-making in a different way? By taking advantage on the wealth of data now available, valuable new insights can be generated that help to rewrite the rulebook on decision-making.
Data analytics can pave the way to valuable new insights to support decision making and address growth analytical trends.
As a concrete showcase, this report outlines the main methodological steps for creating one of the most important solutions in the industry: A credit scoring model. This provides a tool to the decision-maker to assess the likelihood of default of a new client.
The methodology behind it requires looking into the data to identify statistical trends and patterns that can be used in the future. It synthesises analytics with business, to answer the particular client needs.
In this report we emphasise the various ways to assess model performance, goodness-of-fit and predictive power, as well as some typical refinements that help improve it further.
Topics discussed are:
- Data quality testing
- Variable selection
- Logistic regression
- Information Value
- ROC curves
- Goodness of fit
- Model interpretation
The report also illustrates how to extract transparent interpretations out of the model, a holy grail for the success of a model to the business.