A Framework to Cluster Temporal Data Using Personalised Modelling Approach

A review on data clustering using spiking neural network (SNN) models

“…It is also shown that Spike-Temp and Spike-CD can classify diverse data set with better-quality of accurateness and a shorter simulation time is need compared to the rank order method [16]. Moreover, the features representation in this method can be recycled in deep neural networks to excerpt enhanced feature structure [15].…”

“…In this paper, few applications of SNN in data clustering is presented. As an unsupervised learning, clustering tends to organize data samples into it homogeneous groups in which all the data samples within the group are similar constructed on particular designated possessions of the data [15]. Unsupervised clustering removes the requirement for several repetitions over a training data set which is typically obligatory for supervised learning procedures for example gradient lineage and its deviations [16] thus it require lower computational burden.…”

Evolving spiking neural networks methods for classification problem: a case study in flood events risk assessment

“…SVM consists of a kernel function, basically an equation which divides data vectors into different classes based on which area the data vector falls on [2,21] . A model created using global modelling can be easily applied to new data, however, additional knowledge regarding the data such as the nature of the data and knowledgedatabase is naturally neglected in global modelling which causes information loss [3,12] . Due to this characteristic of global modelling, it is not suitable to be used to analyse the dynamic and Local Modelling-introduced to solve problems of global modelling, where local modelling is more adaptable to new data vectors [2,21] .…”

“…Personalised Modelling-created based on a single point from a subset of the whole problem space, in which every new data vectors can be classified into their corresponding classes based on the model which is constructed "on the fly" [3,12,22] . K-Nearest Neighbor (K-NN) is a modelling technique which for every new samples, the nearest K samples are derived from the data set using Euclidean distance measure and a personalized voting which then labels the samples to its appropriate cluster [3] .…”

“…An investigation on natural events requires analysis on complex high dimensional data in the form of spatio-and spectro-temporal data (SSTD) and it is important to retain the interrelationship information between the components. Data mining tools provides a promising framework for discovering hidden patterns within a multi-dimensional dataset, allowing prediction of occurrences-applicable in various domains like prediction of environmental disaster occurrences, risk of stroke [1,3,4,[6][7][8][9], assessment of ecological data [5,12,13] prediction of unknown gene functions [14] , and many others [7,[15][16][17] .…”