A Customised LSTM-Based Deep Learning Framework for Transformer Predictive Maintenance: Performance Analysis | Chapter 5 | New Horizons of Science, Technology and Culture Vol. 9

 

Transformers are critical and costly components of power systems whose health deteriorates over time due to factors such as poor cooling and heavy loading. Consequently, predictive maintenance is emerging as an effective alternative to conventional corrective maintenance, enabling continuous monitoring and early fault detection.

 

To enhance the effectiveness of predictive maintenance for power transformers under limited Dissolved Gas Analysis (DGA) data conditions, this study proposes a customised Long Short-Term Memory (C-LSTM) deep learning model. The developed C-LSTM architecture is specifically designed to address the limitations of conventional LSTM networks, which often exhibit higher classification error rates when trained on small datasets and may underperform compared to traditional machine learning approaches.

 

A comprehensive performance evaluation was conducted by comparing the proposed C-LSTM model with several well-established traditional machine learning algorithms using multiple metrics, including validation accuracy, test accuracy, precision, recall, and F1-score. Additionally, the diagnostic capability of the model was rigorously assessed across seven transformer fault categories, including low- and high-energy discharges, partial discharge, electrical and thermal faults, and low-, medium-, and high-temperature thermal faults.

 

The experimental results demonstrate the superior classification and diagnostic performance of the proposed C-LSTM model, achieving a validation accuracy of 100% and a test accuracy of 98.57%, significantly outperforming conventional machine learning techniques. These findings confirm that the proposed C-LSTM framework offers a robust and reliable solution for transformer fault diagnosis and predictive maintenance, particularly in scenarios characterised by scarce DGA datasets.

 

 

Author(s) Details

G.V.S.S.N. Srirama Sarma
Department of Electrical and Electronics Engineering, Matrusri Engineering College, Saidabad, Hyderabad, India.

 

Please see the book here :- https://doi.org/10.9734/bpi/nhstc/v9/6804

 

A Clinical Decision Support Model for Predicting Avoidable Re-hospitalization of Breast Cancer Patients in Kenyatta National Hospital |Chapter 7 | Mathematics and Computer Science: Contemporary Developments Vol. 2

 

This study develops a Clinical Decision Support Model (DSM) to aid in assessing and recommending the discharge of a breast cancer patient from the hospital ward since the discharge problem is often overwhelming for clinicians to process at the point of care or in urgent situations. The model incorporates breast cancer patient-specific data that is well-structured, having been obtained from pre-study administered questionnaires and current evidence-based guidelines. The obtained dataset of the pre-study questionnaires is processed using data mining techniques to generate an optimal clinical decision tree classifier model. This model assists physicians in enhancing their decision-making process when discharging a patient, based on basic cognitive processes in medical thinking. This resulted in new, better-formed, and superior discharge outcomes. The model improves the quality of patient discharge assessments through a predictive discharging model outcome designed from individual unique risk attributes at the point of discharge. This enables timely detection of possible deterioration in health quality upon said discharge, which is noted as a major contributor to ward congestion as a result of re-hospitalization from poor discharge assessment that currently wholly relies on overwhelmed clinicians at the point of care or in urgent situations. The outcome of the implemented model is that it bridges the gap caused by less informed clinical discharge, and reinforces discharge decisions that ensure better treatment outcomes, thus reducing unforeseeable deterioration in the quality of health for discharged patients and surges in the mortality rate blamed on mistrusted discharge decisions. This paper is organized to start with a discussion of the breast cancer scourge and clinical knowledge for discharging patients, data mining techniques, the classifying model accuracy, and the Python web-based decision support model that predicts avoidable re-hospitalization of a breast cancer patient through an informed clinical discharging support model.

Author(s) Details

Christopher Oyuech Otieno
Department of Computer Science, University of Nairobi, Nairobi, Kenya.

Oboko Robert Obwocha
Department of Computer Science, University of Nairobi, Nairobi, Kenya.

Andrew Mwaura Kahonge
Department of Computer Science, University of Nairobi, Nairobi, Kenya.

 

Please see the book here:- https://doi.org/10.9734/bpi/mcscd/v2/1022

Detection and Selection of Task-specific Features Algorithms for IoT-based Networks | Chapter 6 | Research Updates in Mathematics and Computer Science Vol. 6

 

In IoT-based home/enterprise network applications, an advanced security system is desirable for resource-constrained devices. Feature selection significantly affects the performance of a Machine Learning-based Intrusion Detection System (ML-IDS) to which data of the highest quality should be fed. An appropriate feature selection with sufficient features increases the accuracy of the Intrusion Detection System (IDS) classification. In addition, the consistent use of the same metrics in feature selection and detection algorithms further enhances classification accuracy. First, this paper studies two feature selection algorithms, Information Gain, a metric of entropy, and PSO-based feature selection, a metric of misclassification, to select a minimum number of attack feature subsets for resource-constrained IoT devices. Then, the detection algorithms for multi-classifications, Tree and Ensemble, are evaluated regarding non-consistent and consistent metrics. For specific performance comparison, the same metrics for feature selection and detection algorithm are utilized and compared with non-consistent use of feature selection and detection algorithm, e.g., feature selection by Information Gain (entropy) and Tree detection algorithm by classification.

 

Author(s)details:-

 

Yang Kim
Department of Computer Engineering Technology, City Tech, CUNY, NY 11201, USA.

 

Benito Mendoza
Department of Computer Engineering Technology, City Tech, CUNY, NY 11201, USA.

 

Ohbong Kwon
Department of Computer Engineering Technology, City Tech, CUNY, NY 11201, USA.

 

John Joon
Department of Mathematics, Computer Science and Cybersecurity, Mercy University, Dobbs Ferry, USA.

 

Please See the book here :-  https://doi.org/10.9734/bpi/rumcs/v6/3773G

A Novel Approach to Analyze Pranayama through Machine Learning Techniques | Chapter 07 | New Approaches in Engineering Research Vol. 12

 Pranayama (breathing exercises) is an integral part of yoga practise. When practising pranayama, it's important to keep track of how many cycles you've done. The most important thing is to keep track of how long someone is inhaling and exhaling. Maintaining a proper ratio throughout the inhalation: exhalation cycle is critical. The counting procedure is so taxing for a beginning that it is difficult to retain awareness of the breathing process, and the standard of pranayama practise suffers as a result. The goal of the proposed system is to develop a new method for analysing the quality of Pranayama using machine learning techniques. The main goal of the proposed project is to develop an application that can count inhalations and exhalations. It guarantees that users receive feedback based on their breathing and exhalation patterns. It analyses each breath and exhalation pattern and uses Clustering algorithms to classify inhalation and exhalation. This paper's proposed structure aids in improving the uniformity of pranayama. As a result, respiratory performance improves, which reduces melancholy and anxiety. To test the validity of breathing patterns, the KNN, SVM, Random Forest, and Decision Tree algorithms are used.

Author(S) Details

A. Parkavi
Department of CSE, M S Ramaiah Institute of Technology, Bangalore, India.

V. Sangeetha
Department of CSE, M S Ramaiah Institute of Technology, Bangalore, India.

G. R. Amith
Department of CSE, M S Ramaiah Institute of Technology, Bangalore, India.

B. K. Harini
Department of CSE, M S Ramaiah Institute of Technology, Bangalore, India.

M. Supriya
Department of CSE, M S Ramaiah Institute of Technology, Bangalore, India.

K. N. Tejasvini
Department of CSE, M S Ramaiah Institute of Technology, Bangalore, India.

View Book:- https://stm.bookpi.org/NAER-V12/article/view/3841

A Novel Approach to Analyze Pranayama through Machine Learning Techniques | Chapter 7 | New Approaches in Engineering Research Vol. 12

Pranayama (breathing exercises) is an integral part of yoga practise. When practising pranayama, it's important to keep track of how many cycles you've done. The most important thing is to keep track of how long someone is inhaling and exhaling. Maintaining a proper ratio throughout the inhalation: exhalation cycle is critical. The counting procedure is so taxing for a beginning that it is difficult to retain awareness of the breathing process, and the standard of pranayama practise suffers as a result. The goal of the proposed system is to develop a new method for analysing the quality of Pranayama using machine learning techniques. The main goal of the proposed project is to develop an application that can count inhalations and exhalations. It guarantees that users receive feedback based on their breathing and exhalation patterns. It analyses each breath and exhalation pattern and uses Clustering algorithms to classify inhalation and exhalation. This paper's proposed structure aids in improving the uniformity of pranayama. As a result, respiratory performance improves, which reduces melancholy and anxiety. To test the validity of breathing patterns, the KNN, SVM, Random Forest, and Decision Tree algorithms are used.

Author (S) Details

A. Parkavi
Department of CSE, M S Ramaiah Institute of Technology, Bangalore, India.

V. Sangeetha
Department of CSE, M S Ramaiah Institute of Technology, Bangalore, India.

G. R. Amith
Department of CSE, M S Ramaiah Institute of Technology, Bangalore, India.

B. K. Harini
Department of CSE, M S Ramaiah Institute of Technology, Bangalore, India.

M. Supriya
Department of CSE, M S Ramaiah Institute of Technology, Bangalore, India.

K. N. Tejasvini
Department of CSE, M S Ramaiah Institute of Technology, Bangalore, India.

View Book :- https://stm.bookpi.org/NAER-V12/article/view/3841

Study on Auto Sector Stock Price Trend Prediction by using Decision Tree| Chapter 7 | New Approaches in Engineering Research Vol. 5

 The auto sector stock price trend is influenced by a number of national and worldwide unknown elements. It takes into account both local and global influences. The influence of such a factor on the stock price trend is difficult to forecast because the impact of the same factor fluctuates over time and due to the nonlinear character of the financial stock market. One of the strategies for capturing trends in historical data is machine learning. This feature is what inspired us to apply it in this study to forecast the auto sector stock price trend. It suggests a thorough investigation into the auto sector's stock price trend forecasting. This study focuses on decision tree classifier as a method for forecasting trend in the auto industry price prediction since it can handle multidimensional data and does not require domain knowledge.

 
Author(s) Details

Manish M. Goswami
Department of Information Technology, Rajiv Gandhi College of Engineering & Research, Nagpur, India.

Sachin Jambhulkar
Department of Electronics & Communication, Vidarbha Institute of Technology, Nagpur, India.

View Book :- https://stm.bookpi.org/NAER-V5/article/view/2088

Time Series Variation Based on Power Links and Field Association Words in English and Arabic | Book Publisher International

 Some words have frequency appearance in each text and are known as keywords because they have a close association with the subjects of their texts, these words change frequencies in a given duration with time series variation. However, the significance of frequency shift with time-series variance is not considered in conventional text dealing methods and text search techniques. Traditional methods were therefore unable to correctly determine the popularity index of the word in a given time. A new method is proposed to automatically estimate the stability groups (increasing, relatively constant, and decreasing) that indicate the popularity of the word with variations in time series based on the change in frequency in previous text data. At first, learning data was created by defining five attributes to quantitatively measure the frequency of word shift, automatically extracting these five attributes from electronic texts. These learning data were categorised manually (human) into three classes of stability. These data were then subjected to a decision tree (DT) for automatic determination of the stability groups of research data (test data). However, the accuracy of the decision tree estimate decreases as data numbers are distributed between groups. A new way to use a Random Sampling Approach and a new Data Copying Method to increase the accuracy of decision tree estimation. In addition, the term Field Association (FA) is the smallest word or conceptual unit capable of evaluating a text field in a scheme called field tree. In a given period of time, a modern methodology based on frequencies of particular terms related to the field called Field Association (FA) terms is used. The technique suggests that the stabilisation classes of FA terms are automatically evaluated to point out the popularity of the list of FA terms depending on the time shift and to increase the accuracy of the Decision Tree (DT). In addition, all previous methods are based on keywords in the languages of English and French. The extension of keywords to other languages, such as Arabic, could therefore reinforce further research in that area. This approach incorporates a new method of extracting Arabic keywords from companies based on changes in a text over several periods of time using a decision tree based on their recurrences. The new methodology is applied to the new field of data set (computer science) that differentiates it from historically used approaches. Finally, to examine the co-occurrences of words in the publications of a given topic, Power Links Analysis (PLA) was developed to solve the previous downside of conventional approaches. This approach is focused on both the advanced frequency form and the distance between the various instances of the terms given. A new technique has been proposed that extracts stability groups of field association terms based on automated power relation analysis to improve decision tree accuracy. In this method, we analysed the effects of time variance based on the frequencies of unique FA words that applied to documents in a particular period using power connection.

Author(s) Details

El-Sayed Atlam
Department of Computer Science, Faculty of Science, Tanta University, Tanta, Egypt and College of Computer Science and Engineering at Yanbu, Taibah University, Saudi Arabia.

View Book :- https://bp.bookpi.org/index.php/bpi/catalog/book/312