A study of the effect of different types of noise on the precision of supervised learning techniques

Abstract

Machine learning techniques often have to deal with noisy data, which may affect the accuracy of the resulting data models. Therefore, effectively dealing with noise is a key aspect in supervised learning to obtain reliable models from data. Although several authors have studied the effect of noise for some particular learners, comparisons of its effect among different learners are lacking. In this paper, we address this issue by systematically comparing how different degrees of noise affect four supervised learners that belong to different paradigms. Specifically, we consider the Naïve Bayes probabilistic classifier, the C4.5 decision tree, the IBk instance-based learner and the SMO support vector machine. We have selected four methods which enable us to contrast different learning paradigms, and which are considered to be four of the top ten algorithms in data mining (Yu et al. 2007). We test them on a collection of data sets that are perturbed with noise in the input attributes and noise in the output class. As an initial hypothesis, we assign the techniques to two groups, NB with C4.5 and IBk with SMO, based on their proposed sensitivity to noise, the first group being the least sensitive. The analysis enables us to extract key observations about the effect of different types and degrees of noise on these learning techniques. In general, we find that Naïve Bayes appears as the most robust algorithm, and SMO the least, relative to the other two techniques. However, we find that the underlying empirical behavior of the techniques is more complex, and varies depending on the noise type and the specific data set being processed. In general, noise in the training data set is found to give the most difficulty to the learners.

Disease Diagnosis in Smart Healthcare: Innovation, Technologies and Applications

“…Because there are numerous practical challenges, they cannot be simply treated as black-box (simply enter the input features). Some of the difficulties are: large dimensionality of feature vectors [50], bias/variance dilemma [51], input and output noise [52], large-scale training data [53], data heterogeneity [54], data redundancy [55] and non-linearity among features [56].…”

A survey on pre-processing techniques: Relevant issues in the context of environmental data mining

“…In most applications, each parameter (or variable) will represent a column of the data matrix and the timestamp will be an additional variable. Sometimes, some pre-processing is required to integrate the information of several sensors in a single indicator (like mean water level of a bioreactor for example, measured by several sensors in different points) and the imprecision or uncertainty associated to the measurements must be properly treated at pre-processing level, in particular the noise associated to the signals [168]. Regarding imprecision, georadar, or GPS systems, for example, provide a region where the target can be located with a certain probability, but are not able to provide exact positions of the target instances.…”

Conservative Noise Filters

“…Domingos et al [32] found that NB performance is competitive with more sophisticated ML methods, such as DT, IBL, and rule induction, even if the features' dependency is very strong. Moreover, NB is a strongly noise-tolerant algorithm [33], [34]. Nettleton et al [33] performed a systematic analysis of robustness of many ML algorithms to noise-namely, NB, C4.5, IBk, and SMO.…”