Faculty Publications

Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736

Publications by NITK Faculty

Browse

Has files: NoSubject: search.filters.subject.Time series

Search Results

Now showing 1 - 10 of 21
  • No Thumbnail Available
    Item
    A TFD Approach to Stock Price Prediction
    (Springer, 2020) Chanduka, B.; Bhat, S.S.; Rajput, N.; Mohan, B.R.
    Accurate stock price predictions can help investors take correct decisions about the selling/purchase of stocks. With improvements in data analysis and deep learning algorithms, a variety of approaches has been tried for predicting stock prices. In this paper, we deal with the prediction of stock prices for automobile companies using a novel TFD—Time Series, Financial Ratios, and Deep Learning approach. We then study the results over multiple activation functions for multiple companies and reinforce the viability of the proposed algorithm. © 2020, Springer Nature Singapore Pte Ltd.
  • No Thumbnail Available
    Item
    Deep Learning-Based Prediction, Classification, Clustering Models for Time Series Analysis: A Systematic Review
    (Springer Science and Business Media Deutschland GmbH, 2022) Naik, N.N.; Chandrasekaran, K.; Venkatesan, M.; Prabhavathy, P.
    Analysis of time series is a prominent issue in the field of data analysis. With large amount of existing data in time series, multiple algorithms for analyzing time series data are being proposed. A variety of deep learning models are being designed to enhance the diversity of datasets related to time series across different fields. In comparison with the existing methods, only few have incorporated deep neural networks to perform this task. In most of the cases, deep neural networks are being applied for image data but it can also be used for sequential data such as text and audio. Here, we throw light on the recent advancements in hybrid deep learning models which consist of combination of various frameworks of deep neural networks with statistical models that have led to an improvement in time series analysis. Deep learning models are categorized into discriminative, and generative models provide an insight into the data based on the perception of conditional or joint probability. In this paper, we have surveyed newly devised algorithms and limitations of prediction, classification and clustering for time series analysis which describes how the temporal information can be merged into the analysis of the time series data. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
  • No Thumbnail Available
    Item
    Epidemic Outbreak Prediction with Ensemble of Deep Learning Models
    (Institute of Electrical and Electronics Engineers Inc., 2022) Vasudev, R.; Dahikar, P.; Jain, A.; Patil, N.
    The ability of today's technology has proved it's significance and dire need in the world yet again, with COVID-19 being a global pandemic. Various techniques are being incorporated and researches being conducted everyday in order to mitigate this pandemic. Forecasting of COVID-19 cases is one such task in machine learning which is being researched intensively to develop reliable forecasting models.In the proposed work, we have forecasted the number of COVID-19 confirmed,recovered and death cases globally using time series data with machine learning and deep learning ensemble models. The purpose of this study is to prove that ensemble of several week learners that we have developed can result in a better performing model. Deep learning models always tend to perform better than machine learning and traditional linear models due to their non-linearity. Our study concludes that deep learning ensemble model achieves better performance than the machine learning ensemble (Random forest) and the individual base learners used in ensemble model itself in COVID-19 forecasting. © 2022 IEEE.
  • No Thumbnail Available
    Item
    A comparative study on RBF and NARX based methods for forecasting of groundwater level
    (2011) Dandagala, D.; Deka, P.C.
    Evaluation and forecasting of groundwater levels through time series model (s) helps for the sustainable development of groundwater resources. The focus of the present study is on the application of Radial Basis Function (RBF) and Non Linear auto-regressive with exogenous variable (NARX) data driven models to forecast groundwater level for multiple input scenario's and also multiple lead time. Weekly time series groundwater level data has been used as input and the models are developed to forecast one, two, three, four, five and sixth week ahead. Root mean square error (RMSE) and correlation coefficient (Cc) are used for evaluating the accuracy of the models. Based on the comparison of results, it was found that the RBF models are superior to the NARX models in forecasting groundwater level considering RMSE and Cc. The obtained result indicates that the RBF has high performance and consistent upto fourth week lead time and decaying performance for NARX models. Hence, RBF and NARX have the potential in forecasting groundwater level efficiently for multi step lead time. © 2011 CAFET-INNOVA TECHNICAL SOCIETY. All rights reserved.
  • No Thumbnail Available
    Item
    Hybrid wavelet neural network model for improving forecasting accuracy of time series significant wave height
    (2011) Prahlada, R.; Deka, P.C.
    Forecasting of a time series ocean wave data for various lead times has been attempted using hybrid wavelet-Artificial neural networks (WLNN) approach in this study. To improve the model performance a wavelet transformation is attached prior to a predictor (ANN) and then analysis has been carried out. Here the wavelet transformation is used to decompose the original significant wave height (Hs) data into its sub signals in the form of approximation coefficients and detail coefficients. Further, these coefficients were fed to ANN as inputs and targets and the results obtained from the hybrid model are then reconstructed to obtain the predicted significant wave heights. The predicted results from the proposed model were compared with the single ANN results. From the results, it is concluded that the proposed model is working efficiently for predicting time series data, and also the error observed at the higher lead time was very less as compared to the single ANN. The effect of decomposition level is also analysed in thisstudy and their influence was observed significantly in the higher lead time forecasting. © 2011 CAFET-INNOVA TECHNICAL SOCIETY.
  • No Thumbnail Available
    Item
    Discrete wavelet neural network approach in significant wave height forecasting for multistep lead time
    (2012) Deka, P.C.; Prahlada, R.
    Recently Artificial Neural network (ANN) was extensively used as non-linear inter-extrapolator for ocean wave forecasting as well as other application in ocean engineering. In this current study, the Wavelet transform was hybridised with ANN naming Wavelet Neural Network (WLNN) for significant wave height forecasting near Mangalore, west coast of India, upto 48 h lead time. The main time series of significant wave height data were decomposed to multiresolution time series using discrete wavelet transformations. Then, the multiresolution time series data were used as input of the ANN to forecast the significant wave height at different multistep lead time. It was shown how the proposed model, WLNN, that makes use of multiresolution time series as input, allows for more accurate and consistent predictions with respect to classical ANN models. The proposed wavelet model (WLNN) results revealed that it was better forecasted and consistent than single ANN model because of using multiresolution time series data as inputs. © 2012 Elsevier Ltd. All rights reserved.
  • No Thumbnail Available
    Item
    Discrete wavelet-Ann approach in time series flow forecasting-a case study of Brahmaputra river
    (2012) Deka, P.C.; Haque, L.; Banhatti, A.G.
    This paper deals with the prediction of hydrologic behavior of the runoff for the one of the largest discharge carrier International River, Brahmaputra, located in Assam (India) at the Pandu station, by using daily time unit. The flow regime dominated by high data non-stationary and seasonal irregularity due to Himalayan climate fallout. The influence of data preprocessing through wavelet transforms has been investigated. For this, the main time series of flow data were decomposed to multi resolution time series using discrete wavelet transformations. Then these decomposed data were used as input to Artificial Neural Network (ANN) for multiple lead time flow forecasting. Various types of wavelets were used to evaluate the optimal performance of models developed. The forecasting accuracy of the models has been tested for multiple lead time upto 4 days using different decomposition levels. The performance of the proposed hybrid model has been evaluated based on the performance indices such as root mean square error (RMSE), coefficient of efficiency (CE) and mean relative error (MRE).The results shows the better forecasting accuracy by the proposed combined hybrid model over the single ANN model in hydrological time series forecasting. © 2012 CAFET-INNOVA TECHNICAL SOCIETY.
  • No Thumbnail Available
    Item
    Hydrological effects of land use /land cover changes on stream flow at Gilgel Abay River Basin, Upper Blue Nile, Ethiopia
    (CAFET INNOVA Technical Society 1-2-18/103, Mohini Mansion, Gagan Mahal Road, Domalguda, Hyderabad 500029, 2016) Mulu, A.; Dwarakish, G.S.
    Water is the most important resource for the survival of living things and it is the most essential resource associated with land use/ land cover (LU/LC) changes. Therefore, it is very important to make evaluations of the expected impact on the hydrology and water resources due to expected changes. The main objective of this study is to assess the hydrological effect of land use/ land cover changes on stream flow at GilgelAbay river basin using Precipitation Runoff Modeling System (PRMS) model. System inputs are daily time-series values of precipitation, minimum and maximum air temperature, and parameter files which are generated from GIS Weasel. To identify effect of changes in LU/LC, vegetation type and vegetation density on stream flow, LU/LC, vegetation type and vegetation density data from 1990-2000 and 2001-2010 years were considered. This different period LU/LC, vegetation type and vegetation density with soil data and DEM were given to GIS Weasel to generate different parameters for PRMS model. These generated parameters together with time series data (daily minimum and maximum air temperature, daily precipitation and daily stream flow) feed to PRMS model to simulate stream flow for the years 1993-2000 and 2001-2008. From the time series data, climate changes (daily maximum and minimum temperature and daily precipitations) were kept the same as baseline period (1993-2000). The stream flow of 2001-2008 compared with baseline period (1993-2000) and the effect of LU/LC, vegetation type and vegetation density was identified using calibrated and simulated PRMS model. Hence, as LU/LC, vegetation type and vegetation density changed from 1993-2000 period to 2001-2010 period, stream flow increased from 7.8% (128.4 Mm3) to 25.3% (432 Mm3) and ET decreased from 4.2% (75 Mm3) to 20% (524 Mm3) from baseline period. For the whole simulation periods (2001-2008) stream flow increased by 10.9% (784 Mm3), but ET decreased 6.7% (43 Mm3) related to baseline periods. © 2016 CAFET-INNOVA TECHNICAL SOCIETY. All rights reserved.
  • No Thumbnail Available
    Item
    Wavelet coupled MARS and M5 Model Tree approaches for groundwater level forecasting
    (Elsevier B.V., 2017) Rezaie-Balf, M.; Naganna, S.R.; Ghaemi, A.; Deka, P.C.
    In this study, two different machine learning models, Multivariate Adaptive Regression Splines (MARS) and M5 Model Trees (MT) have been applied to simulate the groundwater level (GWL) fluctuations of three shallow open wells within diverse unconfined aquifers. The Wavelet coupled MARS and MT hybrid models were developed in an attempt to further increase the GWL forecast accuracy. The Discrete Wavelet Transform (DWT) which is particularly effective in dealing with non-stationary time-series data was employed to decompose the input time series into various sub-series components. Historical data of 10 years (August-1996 to July-2006) comprising monthly groundwater level, rainfall, and temperature were used to calibrate and validate the models. The models were calibrated and tested for one, three and six months ahead forecast horizons. The wavelet coupled MARS and MT models were compared with their simple counterpart using standard statistical performance evaluation measures such as Root Mean Square Error (RMSE), Normalized Nash-Sutcliffe Efficiency (NNSE) and Coefficient of Determination (R2). The wavelet coupled MARS and MT models developed using multi-scale input data performed better compared to their simple counterpart and the forecast accuracy of W-MARS models were superior to that of W-MT models. Specifically, the DWT offered a better discrimination of non-linear and non-stationary trends that were present at various scales in the time series of the input variables thus crafting the W-MARS models to provide more accurate GWL forecasts. © 2017 Elsevier B.V.
  • No Thumbnail Available
    Item
    Clumped-MCEM: Inference for multistep transcriptional processes
    (Elsevier Ltd, 2019) Shetty, K.S.; B, A.
    Many biochemical events involve multistep reactions. Among them, an important biological process that involves multistep reaction is the transcriptional process. A widely used approach for simplifying multistep reactions is the delayed reaction method. In this work, we devise a model reduction strategy that represents several OFF states by a single state, accompanied by specifying a time delay for burst frequency. Using this model reduction, we develop Clumped-MCEM which enables simulation and parameter inference. We apply this method to time-series data of endogenous mouse glutaminase promoter, to validate the model assumptions and infer the kinetic parameters. Further, we compare efficiency of Clumped-MCEM with state-of-the-art methods – Bursty MCEM2 and delay Bursty MCEM. Simulation results show that Clumped-MCEM inference is more efficient for time-series data and is able to produce similar numerical accuracy as state-of-the-art methods – Bursty MCEM2 and delay Bursty MCEM in less time. Clumped-MCEM reduces computational cost by 57.58% when compared with Bursty MCEM2 and 32.19% when compared with delay Bursty MCEM. © 2019 Elsevier Ltd