A comparative study of machine learning methods for ordinal classification with absolute and relative information

Ordinal classification with a spectrum of information sources

“…In addition to nearest neighbor methods, also other classical ordinal classification methods can be augmented to deal with both absolute and frequentist relative information. A recently published study by the present authors (Tang et al, 2021a) has demonstrated the effectiveness of augmenting classical ordinal classifiers, such as POM, SVLOR, SVOREX, SVORIM and LDLOR (see Gutiérrez et al, 2015 for a survey on ordinal regression methods), by incorporating additional relative information. Similar to the present work, it would also be natural to investigate how to incorporate frequentist relative information to augment such classifiers.…”

“…For illustrative purposes, here we consider LDLOR and carry out some preliminary experiments. In Tang et al (2021a), it was proposed to translate relative information into constraints in the optimization problem associated with LDLOR. Formally, the constraints were as follows: 𝐰 𝑇 𝐚 𝓁 − 𝐰 𝑇 𝐛 𝓁 ≥ 1 − 𝜂 𝓁 , where 𝐰 is a weighing vector and 𝐚 𝓁 and 𝐛 𝓁 are examples forming a couple of relative information (𝐚 𝓁 being preferred to 𝐛 𝓁 ).…”

Active learning for ordinal classification based on expected cost minimization

“…Therefore, proposing a cost-sensitive AL method to learn a promising ordinal classifier with minimal comprehensive cost is worthwhile. (2) To further reduce the labeling cost, we would like to consider the annotator can provide low-cost instance-pair relation information 11 . Thus, investigating active learning for ordinal classification by querying instance-pair relation information is valuable.…”

“…As a supervised learning task, OC usually relies on a sufficient amount of labeled data to train an ordinal prediction model or induce the rules. However, the label acquisition for ordinal instances is usually expensive and time-consuming due to the dependence on human preference and domain expertise 10 , 11 , prohibiting the collection of a large number of labeled instances. In this situation, one can use the active learning (AL) technique 12 – 14 to train an ordinal classifier 15 , 16 .…”

Long-Term Traffic Forecasting in Optical Networks Using Machine Learning

“…One of the most widely used error function is mean absolute error (MAE) [41]]. It is used in a number of ordinal classification problems [42], [43], [44], [45], [46]. Let 𝑌 * = (𝑦 1 * , 𝑦 2 * , … , 𝑦 𝑛 * ) be the set of classes returned by the algorithm in the ordinal classification task, and 𝑌 = (𝑦 1 , 𝑦 2 , … , 𝑦 𝑛 ) be a set of real classes corresponding to 𝑌 * .…”