The Prediction Advantage: A Universally Meaningful Performance Measure for Classification and Regression

نویسندگان

  • Ran El-Yaniv
  • Yonatan Geifman
  • Yair Wiener
چکیده

We introduce the Prediction Advantage (PA), a novel performance measure for prediction functions under any loss function (e.g., classification or regression). The PA is defined as the performance advantage relative to the Bayesian risk restricted to knowing only the distribution of the labels. We derive the PA for well-known loss functions, including 0/1 loss, cross-entropy loss, absolute loss, and squared loss. In the latter case, the PA is identical to the well-known R-squared measure, widely used in statistics. The use of the PA ensures meaningful quantification of prediction performance, which is not guaranteed, for example, when dealing with noisy imbalanced classification problems. We argue that among several known alternative performance measures, PA is the best (and only) quantity ensuring meaningfulness for all noise and imbalance levels.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

S3PSO: Students’ Performance Prediction Based on Particle Swarm Optimization

Nowadays, new methods are required to take advantage of the rich and extensive gold mine of data given the vast content of data particularly created by educational systems. Data mining algorithms have been used in educational systems especially e-learning systems due to the broad usage of these systems. Providing a model to predict final student results in educational course is a reason for usi...

متن کامل

A Validation Test Naive Bayesian Classification Algorithm and Probit Regression as Prediction Models for Managerial Overconfidence in Iran's Capital Market

Corporate directors are influenced by overconfidence, which is one of the personality traits of individuals; it may take irrational decisions that will have a significant impact on the company's performance in the long run. The purpose of this paper is to validate and compare the Naive Bayesian Classification algorithm and probit regression in the prediction of Management's overconfident at pre...

متن کامل

Prediction of melting points of a diverse chemical set using fuzzy regression tree

The classification and regression trees (CART) possess the advantage of being able to handlelarge data sets and yield readily interpretable models. In spite to these advantages, they are alsorecognized as highly unstable classifiers with respect to minor perturbations in the training data.In the other words methods present high variance. Fuzzy logic brings in an improvement in theseaspects due ...

متن کامل

The Porosity Prediction of One of Iran South Oil Field Carbonate Reservoirs Using Support Vector Regression

Porosity is considered as an important petrophysical parameter in characterizing reservoirs, calculating in-situ oil reserves, and production evaluation. Nowadays, using intelligent techniques has become a popular method for porosity estimation. Support vector machine (SVM) a new intelligent method with a great generalization potential of modeling non-linear relationships has been introduced fo...

متن کامل

An Integrated DEA and Data Mining Approach for Performance Assessment

This paper presents a data envelopment analysis (DEA) model combined with Bootstrapping to assess performance of one of the Data mining Algorithms. We applied a two-step process for performance productivity analysis of insurance branches within a case study. First, using a DEA model, the study analyzes the productivity of eighteen decision-making units (DMUs). Using a Malmquist index, DEA deter...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:
  • CoRR

دوره abs/1705.08499  شماره 

صفحات  -

تاریخ انتشار 2017