Dynamic Nearest Neighbor: An Improved Machine Learning Classifier and Its Application in Finances

Abstract

The presence of machine learning, data mining and related disciplines is increasingly evident in everyday environments. The support for the applications of learning techniques in topics related to economic risk assessment, among other financial topics of interest, is relevant for us as human beings. The content of this paper consists of a proposal of a new supervised learning algorithm and its application in real world datasets related to finance, called D1-NN (Dynamic 1-Nearest Neighbor). The D1-NN performance is competitive against the main state of the art algorithms in solving finance-related problems. The effectiveness of the new D1-NN classifier was compared against five supervised classifiers of the most important approaches (Bayes, nearest neighbors, support vector machines, classifier ensembles, and neural networks), with superior results overall.

A novel ensemble approach for estimating the competency of bank telemarketing

“…Furthermore, by comparing the results of EFO-ANN (i.e., EFO-MLP) in this study and classical MLP in Ref. 59 , the effectiveness of the EFO algorithm in enhancing the MLP model is deduced. In another example, a deep convolutional neural network used by Kim et al 32 achieved 76.70% accuracy and it outperformed several other conventional methods.…”

“…It classified 88.76% of the samples correctly and achieved an AUC of 79.31% which is higher than many conventional methods. For instance, Camacho-Urriolagoitia et al 59 applied various machine learning tools to financial problems. As far as the problem of this study is concerned, six models of Naïve Bayes, Logistic, kNN, SVM, MLP, and AdaBoost could correctly classify 76.80%, 85.76%, 86.25%, 76.63%, 88.01%, and 86.58% of the samples, respectively (see “Bank Additional” results in Table 3 of the cited study).…”

Intelligent credit scoring using deep learning methods

“…Ensembles of algorithms were also efficiently combined and applied to the German Credit Score dataset, such as the incremental learning ensemble method (ILEM) 30 and CF-GA-Ens (clustering with fuzzy assignment-genetic algorithm-ensemble learning) 31 and the novel tree-based overfitting-cautious heterogeneous ensemble model (OCHE). 32 Some of the newer studies also applied novel machine learning techniques to German credit scoring models: step-wise multi-grained augmented gradient boosting decision trees, 33 dynamic 1-nearest neighbor 34 and Cost-sensitive Neural Network Ensemble. 35…”

“…Some of the newer studies also applied novel machine learning techniques to German credit scoring models: step‐wise multi‐grained augmented gradient boosting decision trees, 33 dynamic 1‐nearest neighbor 34 and Cost‐sensitive Neural Network Ensemble 35 …”

Methods and Applications of Data Mining in Business Domains

“…In addition to these papers collectively contributing to the field of data-driven applications and predictive modelling across a variety of domains, they offer specific approaches, such as innovative approaches to optimizing product layouts in supermarkets using sequential pattern mining and optimization algorithms [22]. Furthermore, they introduce a novel supervised learning algorithm for financial risk assessment [23] and explore the development trajectory of radio frequency identification (RFID) applications through academic citation and text mining analysis [24]. Another paper presents a deep learning framework for predicting important trading points in the stock market [25], focusing on high-margin opportunities.…”