Klasyfikator liniowy typu CPL uwzględniający koszty błędów klasyfikacji jako narzędzie prognozowania giełdy
Cost-sensitive cpl linear classifier as a market prediction tool

Summary/Abstract: One kind of data mining – classification – can be used for purpose of predicting changes in market prices. In the simplest scenario we can classify every daily market move as one of two classes: increases or decreases. The standard approach to building a classifier is to optimize correctly classified instances (market moves). However, in the case of predicting the stock market, a better measure of model quality could be a potential profit. This article describes such an approach (cost-sensitive classification) for a linear classifier based on a convex and piecewise-linear penalty function (CPL).

• Details
• Contents