Please use this identifier to cite or link to this item:
http://dr.iiserpune.ac.in:8080/xmlui/handle/123456789/8798
Title: | Predictive Statistical Modelling Approaches for Events in Discrete Space and Time |
Authors: | Sinha, Ankur SAHU, SHUBHANKAR Dept. of Mathematics 20191090 |
Keywords: | Research Subject Categories::MATHEMATICS |
Issue Date: | May-2024 |
Citation: | 72 |
Abstract: | This thesis investigates the use of predictive modelling to improve decision-making for events that are distributed over discrete space and occur in discrete time. We explore three main statistical methods: Kernel Density Estimation (KDE), Bayesian methods, and Maximum Likelihood Estimation (MLE) known for their effectiveness in analyzing data and applicability to continuous and discrete time scenarios. A key contribution is developing an advanced KDE model that features adaptive bandwidth and weighted approaches for more accurate density estimates. We focus our modelling approaches on coal production data from various mines; we demonstrate how these predictive models can address uncertainties and periodic trends in mine output. Accurately forecasting mine production can enhance resource allocation and inventory management. The thesis further extends to prescriptive analytics by incorporating mathematical optimization to allocate coal optimally from mines to power plants. This involves formulating and solving a mathematical model with real-world constraints to find the best allocation solution. By combining predictive and prescriptive analytics, these research findings have the potential to significantly impact the sectors where future events can occur only in discrete space and time, leading to more informed and data-driven decisions. |
URI: | http://dr.iiserpune.ac.in:8080/xmlui/handle/123456789/8798 |
Appears in Collections: | MS THESES |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
20191090_Shubhankar_Sahu_MS_Thesis.pdf | MS Thesis | 7.08 MB | Adobe PDF | View/Open Request a copy |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.