Building Decision-making Indicators Through Network Analysis of Big Data

Tomaselli, Venera; Giuffrida, Giovanni; Gozzo, Simona; Rinaldi, Francesco Mazzeo

doi:10.1007/s11205-020-02363-2

Cited by 2 publications

(1 citation statement)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Quantitative framework combined with qualitative research method is the novelty of the paper. In [13], a Decision-making framework through network analysis of big data has been proposed. The paper's methodology is based on combining Big Data analytics with the analysis of network patterns to build decision-making indicators.…”

Section: Related Workmentioning

confidence: 99%

A new model to improve decision-making indicators of bank customers' based on neural networks and particle swarm optimization algorithm

Asghari¹,

Azgomi

negahi³

2022

Preprint

View full text Add to dashboard Cite

Due to the large volume of customers' banking information, classical algorithms face challenges in extracting knowledge and the decision-making process, so algorithms based on artificial intelligence and data mining can help this area. In this paper, we have presented a new model based on the artificial neural network and particle swarm optimization algorithm to improve bank customers' decision-making indicators. The proposed method is based on artificial neural networks and particle swarm optimization (PSO) algorithm. Due to the large volume of available data, the pure neural network can not achieve proper convergence in this area and suffers divergence or overfit. In this paper, the PSO algorithm is used to optimize neural network weights in extracting knowledge from user's banking data to improve the decision indicators. To evaluate the proposed method, two scenarios of optimized neural network using PSO and pure neural networks are used to make decisions between good and bad customers. Then, to evaluate the efficiency of the proposed method, the parameters of accuracy, Sensitivity, and Specificity for both scenarios were extracted and compared. The evaluations show that the accuracy of the optimized neural network is about 97%. In contrast, this value in the pure neural network is about 94%, indicating the proposed method's better performance. The proposed method also has better performance in the sensitivity parameter and feature selection and avoids overfitting than primary neural networks.

show abstract

Section: Related Workmentioning

confidence: 99%