Statistical techniques for index/scoring development

23 Views Asked by At

I have a data matrix $X\in\mathbb{R}^{n\times p}$ where $n$ is the number of instances (independent observations) and $p$ is the number of features (descriptive variables).

I want to obtain an index/score for each of the $n$ instance and for new instances that are not present in the current data matrix.

I know techniques like PCA to project the data into a 1-dimensional space and get an index in such a way that the variability is maximized. Nevertheless, this technique leads you to a not-so-easy to interpret solution.

I would like to construct and index/score easy to interpret, e.g., such that the resulting values are between 0 and 1 and values near 1 are "good" and values near 0 are "bad".

Any suggestion or reference will be appreciated.