Robust Classification Based on Correlations Between Attributes

Abstract

The existence of noise in the data significantly impacts the accuracy of classification. In this article, we are concerned with the development of novel classification algorithms that can efficiently handle noise. To attain this, we recognize an analogy between k nearest neighbors (kNN) classification and user-based collaborative filtering algorithms, as they both find a neighborhood of similar past data and process its contents to make a prediction about new data. The recent development of item-based collaborative filtering algorithms, which are based on similarities between items instead of transactions, addresses the sensitivity of user-based methods against noise in recommender systems. For this reason, we focus on the item-based paradigm, compared to kNN algorithms, to provide improved robustness against noise for the problem of classification. We propose two new item-based algorithms, which are experimentally evaluated with kNN. Our results show that, in terms of precision, the proposed methods outperform kNN classification by up to 15%, whereas compared to other methods, like the C4.5 system, improvement exceeds 30%.

IVESA - Visual Analysis of Time-Stamped Event Sequences

“…With INFUSE [58], users can also apply interactive feature selection tasks, here to support prediction modeling. Finding features to characterize TSEQs can often be based on work in extracting temporal features from time-series data [78], on applying metrics [92], or both. However, our task to extract features through metrics was considerably impeded by the fact that most inspiring work for time series and classical event sequences takes the value information into account, which does not exist for TSEQs.…”

“…Experts also pointed out gaps, outliers, periodicity, subsequence length, and dense regions as important. The awareness of their relevance also helped us to prioritize metrics identified in the literature: Related works come from the domain of statistical metrics for time-series [78], where we identified a subset of metrics, applicable for TSEQs. Summary metrics such as the number of events or minimum, maximum, and mean length of sequences form a source for features at the TSEQs granularity.…”

A Lightweight Data Preprocessing Strategy with Fast Contradiction Analysis for Incremental Classifier Learning

“…The advantages include simplicity in merely computing the support and confidence values for estimating which target label one instance should be classified into, no persistent tree structure or trained model needs to be retained except small registers for statistics, and the samples (reference) required for noise detection can scale flexibly to any amount (≤ ). One example that is inspired by [27] about a weighted PWC is shown in Figure 4.…”

Underwater Sonar Signals Recognition by Incremental Data Stream Mining with Conflict Analysis

“…The advantages include simplicity in merely computing the supports and confidence values for estimating which target label one instance should be classified into, no persistent tree structure or trained model needs to be retained except small registers for statistics, and the samples (reference) required for noise detection can scale flexibly to any amount (≤ W ). One example that is based on [17] about a weighted PWC is shown in Figure 3.…”