Bank transaction data modeling by optimized hybrid machine learning merged with ARIMA

Abstract

The bank transactions are needed to be modeled to predict the future transactions of the banks based on the previous transactions. In order to achieve efficient modeling of bank data transactions, Deep Belief Network (DBN) and Neural network (NN) classifiers are used in this paper. Initially, the bank transaction data such as transaction count and amount are subjected to feature extraction to extract the statistical features. Now, the extracted data are modeled using the combination of DBN and NN models, where the average modeled output from both the network is considered as the final result. The above procedure is utilized for the two prediction models such as transaction count and transaction amount. Moreover, the transaction count from prediction model 1 is subjected to the Auto-Regressive Integrated Moving Average (ARIMA) model to compute the relationship between the transition count and transition amount. Here, as the main contribution, the number of hidden neurons in both DBN and NN are optimized or tuned accurately using the hybridized optimization models with Lion Algorithm (LA), and Artificial Bee Colony (ABC) named L-ABC model. The average of entire transactional amounts, i.e. the modeled outputs are matched with the actual data to validate the performance of the implemented model.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 25 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.