perusalgoritmeista

Perusalgoritmeista, or perusal algorithms, is a class of algorithms used to break down data into more manageable pieces. The concept is related to the idea of exploratory data analysis (EDA), where initial data is not immediately fed into a more complex algorithm but first looked at to pre-evaluate quality.

The name "perusal algorithm" comes from the idea of giving a brief glance over, or perusal, rather than a deeper dive. These algorithms often prove useful in large datasets, where the beginning stages of analysis involve inspecting and exploring the data for correctness or validity.

Perusal algorithms serve as an initial step in data cleaning or preparation, refining or editing data prior to sending it through more complex models. Their goal is not to evaluate the final model or predictive accuracy, but rather to pre-qualify data for further use. Examples of perusal algorithms include inspecting spreadsheet columns, evaluating data consistency, and identifying potential anomalies.

Ranjan Maitra in his 2000 paper "Perusal Algorithms: Not Your Grandfather's Preprocessing Algorithms" defined perusal algorithms as: "Perusal is the unsystematic process of examining an object for defects or suitability for further use; that is, it's a cursory glance."

While the primary use of perusal algorithms is in preliminary data analysis, and smoothing initial data so that it is better suited for feeding into modelling algorithms, they can indeed serve as utility in some Data Mining and data analysis contexts.