This paper compares machine learning methods applied to a difficult real-world problem: predicting computer hard-drive failure using attributes monitored internally by individual drives. The problem is one of detecting rare events in a time series of noisy and nonparametrically-distributed data. The paper develops a new algorithm based on the multiple-instance learning framework and the naive Bayesian classifier (mi-NB) which is specifically designed for the low false-alarm case, and is shown to have promising performance. Other methods compared are Support Vector Machines (SVMs), unsupervised clustering, and non-parametric statistical tests (rank-sum and reverse arrangements). The failure-prediction performance of the SVM, rank-sum and mi-NB algorithm is considerably better than the threshold method currently implemented in drives, while maintaining low false alarm rates.
Did you find this white paper useful?
20 out of 50 users found this white paper useful
Related white papers
TechNet Webcast: Optical Media in Windows Vista (Level 300)
This webcast provides an overview of some of the enhancements of using optical media in the Windows Vista operating system. It covers the new user interface for backing up and...
Case Study: TRIZ Solves a Hard Drive Reliability Problem
TRIZ is the Russian acronym for The Theory of Inventive Problem Solving. TRIZ is a unique knowledge-based technology for accelerated development of design concepts. The drive needs to be protected...
Hard Drive Recovery Case Study: Power Failure Surge Brown Out
Hard drive recovery from electrical issues requires special tools and training. Hard drives these days require more advanced techniques to solve electronic hard drive failures. When a surged disk comes...
Selecting Hard Drives for Test, Measurement, and Control Systems
The majority of today's test, measurement, and control systems include at least one hard drive. This hard drive may reside in the desktop or laptop PC that is controlling the...
Analysis of USB Flash Drives in a Virtual Environment
This paper is a continuation of the authors previous work in which the author proposed an approach where two environments, conventional and virtual, are used independently in the forensic analysis...
PC Gamers Know It's What Inside That Counts
Falcon Northwest has been a leader in the PC gaming world since it opened its doors in 1992. Falcon Northwest to offer gamers more than one option when choosing components,...
University of New Hampshire's 'Project 54' Helps Keep Troopers' Eyes on the Road With Help From Hitachi Endurastar Hard Drives
Distracted drivers may be the butt of many a popular joke - simultaneously chatting on cell phones, fiddling with CD players and drinking lattes while steering their cars through rush-hour...
