Experimental Analysis of Hyperparameters for Deep Learning-Based Churn Prediction in the Banking Sector

Until recently, traditional machine learning techniques (TMLTs) such as multilayer perceptrons (MLPs) and support vector machines (SVMs) have been used successfully for churn prediction, but with significant efforts expended on the configuration of the training parameters. The selection of the right training parameters for supervised learning is almost always experimentally determined in an ad hoc manner. Deep neural networks (DNNs) have shown significant predictive strength over TMLTs when used for churn predictions. However, the more complex architecture of DNNs and their capacity to process huge amounts of non-linear input data demand more time and effort to configure the training hyperparameters for DNNs during churn modeling. This makes the process more challenging for inexperienced machine learning practitioners and researchers. So far, limited research has been done to establish the effects of different hyperparameters on the performance of DNNs during churn prediction. There is a lack of empirically derived heuristic knowledge to guide the selection of hyperparameters when DNNs are used for churn modeling. This paper presents an experimental analysis of the effects of different hyperparameters when DNNs are used for churn prediction in the banking sector. The results from three experiments revealed that the deep neural network (DNN) model performed better than the MLP when a rectifier function was used for activation in the hidden layers and a sigmoid function was used in the output layer. The performance of the DNN was better when the batch size was smaller than the size of the test set data, while the RemsProp training algorithm had better accuracy when compared with the stochastic gradient descent (SGD), Adam, AdaGrad, Adadelta, and AdaMax algorithms. The study provides heuristic knowledge that could guide researchers and practitioners in machine learning-based churn prediction from the tabular data for customer relationship management in the banking sector when DNNs are used.

Heterogenic Autotuner for Electro-Pneumatic Single-Action Actuators

The main concept is to design the novel autotuner in a way that it will introduce benefits that arise from the effect of the fusion of the quantitative and qualitative knowledge gained from identification experiments, long-time expertise, and theoretical findings. The novelty of this approach is in the manner in which the expert heuristic knowledge is used for the development of an easy-to-use and time-efficient tuning process. In the proposed approach, the positioner simply learns, mimics, and follows up the tuning process that is performed by an experienced human operator. The major strength of this approach is that all parameters of positioner PID controller can be estimated by only identifying one single parameter that is the effective time constant of the pneumatic actuator. The elaborated autotuning algorithm is experimentally examined with different commercially available pneumatic actuators and control valves. The obtained results demonstrate that the proposed autotuning approach exhibits good performance, usability, and robustness. This should be considered as particularly relevant in the processes of installing, commissioning, and servicing single-action final control elements.

Assessment of microcredit borrowers by the fuzzy maximin convolution method

The problem of timely repayment of loans at all times has been and continues to be actual for commercial banks. Overcoming this problem substantially depends on the quality of the solvency assessment of potential borrowers, which is carried out by experts on the basis of retrospective information. In the microcredit system, the assessment of the borrower's credit history is usually carried out by an expert who mainly relies on his heuristic knowledge and intuition, which usually extols subjective considerations that do not have sufficient grounds. In practice, the opinions of different analysts or those responsible for making credit decisions often differ, especially if controversial situations are considered that have many acceptable alternative solutions. As a result, in assessing the solvency of potential microloan borrowers, the subjective opinion of the expert and the incompetent or deliberate interpretation of the information resulting in the adoption of decisions that are detrimental to the microfinance organization are overweight. To increase the degree of objectivity, the paper discusses an approach to assessing the responsibility and solvency of microloan borrowers, based on the use of the fuzzy method of maximin convolution. This approach, given the poorly structured personal data of applicants, allows them to be flexibly and quickly assessed for the provision of microloans. The qualitative assessment criteria applied in this case are weighed based on expert opinions regarding the priority of each of them. An important advantage of the proposed model is that it is simple, convenient to use and able to adapt to the requirements of various micro-financial organizations.