Rahul Vyas

Bio: Rahul Vyas is an academic researcher. The author has contributed to research in topics: Artificial neural network & Computational intelligence. The author has an hindex of 1, co-authored 4 publications receiving 1 citations.

A Wilcoxon Norm Based Robust Machine Learning Approach for Traffic Noise Prediction

Abstract: The preliminary objective of this present research work is to construct an empirical traffic noise prediction model for evaluation of equivalent noise level (Leq) in terms of equivalent traffic volume number under heterogeneous traffic flow. For this research work, commercial road networks are preferred for monitoring and modeling. This proposed system introduces a novel method of robust application of wilcoxon norm based machine learning approach (WNN) for traffic noise prediction. The proposed WNN is designed by assuming that training samples used contains strong outliers (high percentage of data corrupt) and the cost function select is a robust norm called Wilcoxon norm. With the presence of outlier most of all computational intelligence models are failure to predict output. In this paper, it is highlights how Wilcoxon norm based artificial neural network model(WNN) has best performance with the presence of outlier compare to conventional multilayer perceptron neural network. For validation, traffic noise problem is consider as a system identification problem at here. From the simulation study it is found that Wilcoxon norm based artificial neural network model has best performance with the presence of outlier

Forecasting Foreign Exchange Rate Using Robust Lagueree Neural Network

Abstract: In this article, single layer based functional basis neural network has been used for foreign exchange rate prediction. In general, foreign exchange rate problem is one of the most complex problems with high non linearity and data irregularity. From many studies it is found that foreign exchange rate prediction always fluctuates with economic growth, interest rate and influence rates and therefore it is very difficult for researcher to predict foreign exchange rate. Therefore, foreign exchange rate prediction becomes a challenging task for every researcher for both academic and industrial communities. In this article two type of single layer functional link artificial neural network Functional-link Artificial Neural Network (FLANN) and Laguerre Polynomial Equation ( LAPE) were applied to forecast foreign exchange data. With high data irregularity, FLANN and LAPE both the models provide extremely precise outcome for complex time series model. The single layered based functional basis neural network architectures results matched strongly with ARIMA with very less Mean Square Error (MSE). From the Simulation study, single layer based functional basis neural network models provide improved results compare to ARIMA model with less Root Mean Square Error (RMSE) and performs as universal approximator.

Predicting Inactiveness in Telecom (Prepaid) Sector: A Complex Bigdata Application

Abstract: A business company especially into telecom operation, suffers from high acquisition cost on new customer rather retaining the in-house customers. As a consequence larger business groups are now spending on retaining those customer who are at the verge of moving out of the service. Even retention activity also accounts for larger portion of the expenditure. In response to these issue, this paper oriented towards finding the ways and means to deriving higher accuracy model along with precision and recall measure of actual inactivity individuals with help of derived KPI's (feature engineering). Various Churn model techniques have been evolved in recent past for the above requirements. The focus of this paper is to manifesting new techniques on feature deriving to unearth hidden pattern on customer behavior, which in-turn helps to determine the Inactive/Churn customer at the higher precision rate.

A Recommendation of Right Product to Right Customer, Next Best Offer: A Comparative Study

Abstract: Telecom companies offer a variety of services and products to stay relevant in the competing market. However, with the vast bouquet of products/services, it becomes difficult for the customer to choose the best-fit product. The current research analyzed two recommendation frameworks i.e., An Adaptive Cognizance (AC) distance based algorithms (i.e. Cosine, Euclidean and Manhattan) and Bayesian Network (BN). Various experiments were carried out to evaluate the performance of AC and BN by comparing them with the existing recommendation system by the Telecom service provider (ES). Both AC and BN could uplift the revenue and conversion in the short term by more than 15%. Interestingly, top 10 impacting recharge products were found to be low cost and, with less validity, promising the rise in long-term revenue owing to increased daily-engagement with the customers by providing next best offers to the telecom prepaid customer.

Noise Impact Assessment and Prediction in MinesUsing Soft Computing Techniques

Abstract: Mining of minerals necessitates use of heavy energy intensive machineries and equipment leading to miners to be exposed to high noise levels. Prolonged exposure of miners to the high levels of noise can cause noise induced hearing loss besides several non-auditory health effects. Hence, in order to improve the environmental condition in work place, it is of utmost importance to develop appropriate noise prediction model for ensuring the accurate status of noise levels from various surface mining machineries. The measurement of sound pressure level (SPL) using sound measuring devices is not accurate due to instrumental error, attenuation due to geometrical aberration, atmospheric attenuation etc. Some of the popular frequency dependent noise prediction models e.g. ISO 9613- 2, ENM, CONCAWE and non-frequency based noise prediction model e.g. VDI-2714 have been applied in mining and allied industries. These models are used to predict the machineries noise by considering all the attenuation factors. Amongst above mathematical models, VDI-2714 is simplest noise prediction model as it is independent from frequency domain. From literature review, it was found that VDI-2714 gives noise prediction in dB (A) not in 1/1 or 1/3 octave bands as compared to other prediction models e.g. ISO-9613-2, CONCAWE, OCMA, and ENM etc. Compared to VDI-2714 noise prediction model, frequency dependent models are mathematically complex to use. All the noise prediction models treat noise as a function of distance, sound power level (SWL), different forms of attenuations such as geometrical absorptions, barrier effects, ground topography, etc. Generally, these parameters are measured in the mines and best fitting models are applied to predict noise. Mathematical models are generally complex and cannot be implemented in real time systems. Additionally, they fail to predict the future parameters from current and past measurements. To overcome these limitations, in this work, soft-computing models have been used. It has been seen that noise prediction is a non-stationary process and soft-computing techniques have been tested for non-stationary time-series prediction for nearly two decades. Considering successful application of soft-computing models in complex engineering problems, in this thesis work, soft-computing system based noise prediction models were developed for predicting far field noise levels due to operation of specific set of mining machinery. Soft Computing models: Fuzzy Inference System (Mamdani and Takagi Sugeno Kang (T-S-K) fuzzy inference systems), MLP (multi layer perceptron or back propagation neural network), RBF (radial basis function) and Adaptive network-based fuzzy inference systems (ANFIS) were used to predict the machinery noise in two opencast mines. The proposed soft-computing based noise prediction models were designed for both frequency and non-frequency based noise prediction models. After successful application of all proposed soft-computing models, comparitive studies were made considering Root Mean Square Error (RMSE) as the performance parameter. It was observed that proposed soft-computing models give good prediction results with accuracy. However, ANFIS model gives better noise prediction with better accuracy than other proposed soft-computing models.

Strategies and Implications of Noise Pollution Monitoring, Modelling, and Mitigation in Urban Cities

Abstract: The present chapter provides an exhaustive review on the noise monitoring studies, comparison of the prediction models including physical propagation model, and applications of the artificial intelligence techniques, noise mapping, and noise pollution monitoring in mining sector carried out by various researchers. Most of the noise pollution studies deal with the assessment of traffic noise and some were focused exclusively on noise monitoring for the residential, educational, industrial, and commercial sites noise. The study reveals that early models were based on mathematical prediction models, later machine learning and deep learning methods were generally used for prediction and forecasting of noise levels. A retrospective view on noise mapping and control is presented in the chapter. Also, the noise pollution control and abatement measures are highlighted that shall be indispensable for reducing the ambient noise levels in metropolitan cities of the country.

Telco customer top‐ups: Stream‐based multi‐target regression

Abstract: Telecommunication operators compete not only for new clients, but, above all, to maintain current ones. The modelling and prediction of the top‐up behaviour of prepaid mobile subscribers allows operators to anticipate customer intentions and implement measures to strengthen customer relationship. This research explores a data set from a Portuguese operator, comprising 30 months of top‐up events, to predict the top‐up monthly frequency and average value of prepaid subscribers using offline and online multi‐target regression algorithms. The offline techniques adopt a monthly sliding window, whereas the online techniques use an event sliding window. Experiments were performed to determine the most promising set of features, analyse the accuracy of the offline and online regressors and the impact of sliding window dimension. The results show that online regression outperforms the offline counterparts. The best accuracy was achieved with adaptive model rules and a sliding window of 500,000 events (approximately 5 months). Finally, the predicted top‐up monthly frequency and average value of each subscriber were converted to individual date and value intervals, which can be used by the operator to identify early signs of subscriber disengagement and immediately take pre‐emptive measures.