Unsupervised Ranking of Numerical Observations based on Magnetic Properties and Correlation Coefficient

Abstract

This paper addresses a novel unsupervised algorithm to rank numerical observations which is important in many applications in computer science, especially in information retrieval (IR). The proposed algorithm shows how correlation coefficients between attribute values and the concept of magnetic properties can be explored to rank multi-attribute numerical objects. One of the main reasons of using correlation coefficients between attribute values and the concept of magnetic properties is that they are easy to compute and interpret. Our proposed Unsupervised Ranking using Magnetic properties and Correlation coefficient (URMC) algorithm can use some or all the numerical attributes of objects and can also handle objects with missing attribute values. The proposed algorithm overcomes a major limitation of the state-of-the-art technique while achieving excellent results.

Binary Classification with Minimum Observations

“…Given the attribute values of a set of objects, we normalize and compute the correlation coefficient scores for each pair of attributes for all objects. Then we use a modified version of Unsupervised Ranking using Magnetic properties and Correlation coefficient (i.e., URMC) algorithm [11] that takes sorted correlation coefficient scores for each pair of attributes from the dataset as an input and returns the weight for each attribute. These attributes' weights are used to rank the objects.…”

“…Alattas et al [11] propose a new ranking algorithm that is inspired by magnetic properties (MP) and clustering. This algorithm groups the attributes into two clusters (i.e., positive and negative cluster) and place each attribute a weight by using Pearson r correlation coefficient.…”

“…Ranking principal curve (RPC)) [12] model is another ranking algorithm that learns a one-dimensional manifold function using five meta-rules to perform unsupervised ranking tasks on multi-attribute observations. The other work that is closely related to [11] is Li et al [13], which is based on a two-phase attribute selection procedure. The first phase, Spearman Ranking Correlation Coefficients (SRCC), identifies irrelevant attributes that can adversely affect the ranking, and the second phase uses Extended Fourier Amplitude Sensitivity Test (EFAST) that presents the total effect for each attribute to ranking and then selects the attributes based on those phases.…”

“…Most state-of-the-art methods for ranking numerical objects remove irrelevant attributes and objects with missing values. Alattas et al's [11] algorithm can deal with all attributes, so there is no need to remove the irrelevant ones.…”

I Know Therefore I Score: Label-Free Crafting of Scoring Functions using Constraints Based on Domain Expertise

“…When it comes to unsupervised or weakly supervised domain for ranking, prior works like (Dehghani et al 2017;Xu et al 2019;Cornuéjols and Martin 2011) usually aggregate opinions from multiple weak supervision signals of ranks to perform ranking. (Alattas et al 2019) is an unsupervised scoring technique that calculates the weights of each feature based on the similarity in their correlation coefficients (feature clustering). Another recently proposed unsupervised scoring technique (Ichikawa and Tamano 2020) is applicable only for binary features.…”