TY - JOUR
T1 - Deep Learning-Based Mathematical Modelling For Predictive Analysis in Media Consumer Behaviour
AU - Elamin, Abd Elmotaleb A.M.A.
N1 - Publisher Copyright:
© 2024, NSP Natural Sciences Publishing Cor. All rights reserved.
PY - 2024
Y1 - 2024
N2 - Advanced predictive models are required to understand and predict consumer behavior due to the rapid evolution of media consumption patterns. This work aims to improve the accuracy of predictive analyses in media consumer behavior by introducing a new method as Bayesian optimized Long-Short Term Memory (LSTM)-Based Mathematical Modelling. The proposed model uses Bayesian optimization techniques to improve performance in capturing temporal dependencies within media consumption data by optimizing LSTM networks for hyperparameter tuning. Models based on long short-term dependencies in sequential data are based on recurrent neural networks, a class of networks well-known for this capacity. To ensure that the LSTM model is precisely tuned to the particular features of media consumption datasets, the Bayesian optimization framework makes it easier to tune hyperparameters automatically. A more accurate and efficient representation of the complex patterns present in media consumer behavior is made possible by combining LSTM networks and Bayesian optimization. The mathematical model based on Bayesian optimized LSTM is increased the accuracy with 99%, which is 9.62% higher than the accuracy of Random Forests, RNN Based Click Stream Model and Gradient Tree Boosting Method. In an era of constant technological and content evolution, the results of this work adds to the growing field of predictive analytics by providing a potent tool for comprehending and forecasting the dynamic nature of media consumer behavior.
AB - Advanced predictive models are required to understand and predict consumer behavior due to the rapid evolution of media consumption patterns. This work aims to improve the accuracy of predictive analyses in media consumer behavior by introducing a new method as Bayesian optimized Long-Short Term Memory (LSTM)-Based Mathematical Modelling. The proposed model uses Bayesian optimization techniques to improve performance in capturing temporal dependencies within media consumption data by optimizing LSTM networks for hyperparameter tuning. Models based on long short-term dependencies in sequential data are based on recurrent neural networks, a class of networks well-known for this capacity. To ensure that the LSTM model is precisely tuned to the particular features of media consumption datasets, the Bayesian optimization framework makes it easier to tune hyperparameters automatically. A more accurate and efficient representation of the complex patterns present in media consumer behavior is made possible by combining LSTM networks and Bayesian optimization. The mathematical model based on Bayesian optimized LSTM is increased the accuracy with 99%, which is 9.62% higher than the accuracy of Random Forests, RNN Based Click Stream Model and Gradient Tree Boosting Method. In an era of constant technological and content evolution, the results of this work adds to the growing field of predictive analytics by providing a potent tool for comprehending and forecasting the dynamic nature of media consumer behavior.
KW - Bayesian Optimization
KW - Hyperparameter Tuning
KW - Long Short-Term Memory
KW - Media Consumer Behavior
KW - Predictive Analysis
KW - Recurrent Neural Networks
UR - https://www.scopus.com/pages/publications/85185516553
U2 - 10.18576/amis/180217
DO - 10.18576/amis/180217
M3 - Article
AN - SCOPUS:85185516553
SN - 1935-0090
VL - 18
SP - 433
EP - 443
JO - Applied Mathematics and Information Sciences
JF - Applied Mathematics and Information Sciences
IS - 2
ER -