Bulletin of Electrical Engineering and Informatics
Not a member yet
2885 research outputs found
Sort by
Development a decision support system for selection healthcare chatbot
Get PDFIt’s an increasing number of healthcare in many countries. Healthcare chatbot can save money, time and meet patient satisfaction. The healthcare would like to select the best or optimal healthcare chatbot but in the real situations, some healthcare may select the healthcare chatbot by own opinions in the organization with several criteria. The purpose of this research is to design and develop a decision support system (DSS) to select healthcare chatbot under the criteria of: i) functionalities; ii) multilingual ability; iii) usability; and iv) security and privacy. According to this research, it can help healthcare to make a reliable decision. The DSS allows users to select the most suitable alternatives of chatbot. The DSS is analyzed by using analytic hierarchy process (AHP). The result show that the DSS was designed to help in complex decision making and show the making decision of decision maker in the reliable and accurate decision. The result found that it is an appropriate technique for using in the DSS to select the suitable healthcare chatbot in accordance with overall criteria effectively including the sensitivity analysis
Enhancing costumer churn prediction with stacking ensemble and stratified k-fold
Get PDFIn the era of rapid technological advancement, the telecommunications industry undergoes significant changes. Factors such as the speed of technological change, high customer expectations, and changing preferences are the main obstacles that affect the dynamics of telecommunications companies. One major issue faced is the high customer churn rate, adversely impacting company revenue and profitability. Previous studies indicate that customer churn prediction remains complex in the telecommunications industry, with opportunities to optimize algorithm selection and prediction model construction methods. This research aims to improve the accuracy of customer churn prediction by employing a complex model that utilizes stacking ensemble learning techniques. The proposed model combines 6 base algorithms: extreme gradient boosting (XGBoost), random forest, light gradient boosting machine (LightGBM), support vector machine (SVM), K-nearest neighbor (KNN), and neural network (NN), with XGBoost as the meta-learner model. The research process involves preprocessing, class data balance with synthetic minority oversampling technique (SMOTE), training using stratified k-fold, and model evaluation. The model is tested using the Telecom Churn dataset. The evaluation results show that the constructed stacking model achieves 98% accuracy, 98.74% recall, 98.03% precision, and 98.38% F1 score. This study demonstrates that optimizing the stacking ensemble model with SMOTE and stratified k-fold enhances customer churn prediction accuracy
Non-centroid-based discrete differential evolution for data clustering
Get PDFData clustering can find similarities and hidden patterns within data. Given a predefined number of groups, most partitional clustering algorithms use representative centers to determine their corresponding clusters. These algorithms, such as K-means and optimization-based algorithms, create and update centroids to give (hyper) spherical shape clusters. This research proposes a non-centroid-based discrete differential evolution (NCDDE) algorithm to solve clustering problems and provide non-spherical shape clusters. The algorithm directs the population of discrete vectors to search for data group labels. It uses a novel discrete mutation strategy analogous to the continuous mutation in classical differential evolution. It also combines a sorting mutation to enhance convergence speed. The algorithm adaptively selects crossover rates in high and low ranges. We use the UCI datasets to compare the NCDDE with other continuous centroid-based algorithms by intra-cluster distance and clustering accuracy. The results show that NCDDE outperforms the compared algorithms overall by intra-cluster distance and achieves the best accuracy for several datasets
Chronic disease prediction chatbot using deep learning and machine learning algorithms
Get PDFEver since the rise of human civilization, more and more diseases have been discovered with the rapid growth of medical knowledge. This sheer volume of information makes it hard for humans to memorize or even utilize it efficiently. Thus, machine learning emerged as a powerful tool for complex calculations by offering a solution to this challenge. This paper intends to use deep learning and machine learning algorithms to develop a predictive model that can recognize potential diseases based on symptoms. The model is then seamlessly integrated into a text-based disease prediction assistant chatbot that serves as a communication platform between the users and the system. The algorithms researched for the disease prediction models are k-nearest neighbours (KNN), support vector machines (SVM), random forest, and neural networks. After that, a chatbot application is created by integrating long short-term memory (LSTM), natural language toolkit (NLTK) libraries, and Telegram. As a result, the SVM models demonstrated excellent performance by achieving an accuracy of 92.24%, closely followed by random forest with 92.23%, KNN with 91.57%, and artificial neural network (ANN) with 91.52% accuracy. In short, this paper presents a potential solution for a more accurate disease prediction tool by implementing the best disease prediction model with the chatbot models together
An opinionated sentiment analysis using a rule-based method
Get PDFThe categorization of opinions into positive, negative, or neutral facilitates information gathering, pinpointing individual weaknesses, and streamlining the decision-making process. Precision in opinion classification enables decision-makers to extract valuable insights, make well-informed decisions, and execute suitable actions. Sentiment analysis is language-specific due to the distinct morphological structures unique to each language, distinguishing them from one another. This study implemented a rule-based sentiment analysis approach for Kafi-noonoo opinionated texts, leveraging a rule-based system tailored for smaller datasets that operate based on a predefined set of rules. The rule-based mechanism calculates the overall polarity of a given sentence by applying a set of rules and categorizes it into positive, negative, or neutral sentiments upon identifying sentimental terms from a dedicated file. While the analysis utilized 1,500 words sourced from Facebook and music review samples, the modest sample size yielded satisfactory results. Performance evaluation metrics such as precision, recall, and F-measure were employed, indicating positive word scores of 91%, 86%, and 88.4%, and negative word scores of 80%, 75%, and 77%, respectively
Application of feature-based image matching method as an object recognition method
Get PDFIn everyday life, objects are recognized based on the suitability of their characteristics to familiar objects. A feature matching process occurs when recognizing objects. This concept is what we want to apply and test in this research. Because various factors can influence the level of accuracy and success of an image matching method, the first step taken is to improve the accuracy level of the image matching method used. There are three feature-based image matching methods, which are implemented as object recognition methods. These three methods are the result of modifications of the image matching function method, normalized 2D cross correlation method and point feature matching which were later named PICMatch, NCMatch and FBMatch. As image matching methods, these three modified methods show performance with a success rate above 95%. However, when applied as an object recognition method, both individually and combined, the three methods only have a maximum accuracy of 7%. These results are obtained by matching the samples using one of the methods with the best match rate, in the order of application of the PICMatch, NCMatch, and FBMatch methods
Optimizing plant health monitoring: improved accuracy and the computational efficiency with stacked machine learning models and feature filtering
Get PDFPlant cultivation can be effectively achieved with the help of hydroponic farming that allows growing soilless and organic plant veggies. However, maintaining optimal plant health in such controlled environments requires continuous monitoring and assessment techniques. This paper provides a comprehensive description of how to determine and categorize the health of hydroponic plants based on a wide range of parameters, such as temperature, pH, electrical conductivity (EC), leaf count, plant height, and vegetative indices. We present a novel approach termed “Hybrid XGBoosting†that combines the multi-classification algorithm extreme gradient boosting (XGBoost) with gradient-based one-side sampling (GOSS) methods to improve accuracy and processing efficiency. This approach first adopts a feature correlation method known as “Pearson’s correlation†for reducing repeated data that are directly proportional or inversely proportional to each other. Finally, we perform a thorough comparative study using well-known algorithms including traditional XGBoost, AdaBoost, and gradient boosting. We demonstrate the better prediction capabilities of Hybrid XGBoosting with 97.93% accuracy through rigorous testing and evaluation, showing its potential for improving hydroponic plant health assessment approaches. Additionally, our research employs comprehensive algorithm assessment measures, such as root mean squared scaled error (RMSSEE), to guarantee the stability and reliability of the results
A 0.7 GHz and 0.9 GHz efficient and compact dual-band rectifier for ambient radio frequency energy harvesting
Get PDFThis study introduces a compact dual-band rectifier utilizing a single and multi-stub matching network (MN) technique. The rectifier consists of two branches, each incorporating a single block stub and two blocks stub to generate two frequency susceptance blocks, subsequently transformed into a meandered line. The proposed rectifier operates at two frequency bands of 0.7 GHz and 0.9 GHz and is fabricated on an RT/Duroid 5880 printed circuit board (PCB) with dimensions of 37×25×1.6 mm using an entire ground architecture. Simulation and measurement results show that the rectifier has a power conversion efficiency (PCE) of 67.77% and 66.35% at 0.7 GHz and 70.31% and 71.22% at 0.9 GHz with input power of 0 dBm, respectively. The rectified voltage is 1.79 V DC across a 5 kΩ load terminal (RL) with 5 dBm input power and is capable of sensing low input power down to -30 dBm. This feature makes the rectifier a promising solution for powering low-power devices from ambient energy
Forward and optimised IK utilising the Levenberg-Marquardt algorithm 6-DOF robot manipulator for sorting applications
Get PDFColour sorting robots automate industries, but translating image data to robot movement is expensive and complicated. Vision sensors require a lot of processing power, which can slow down and strain the robot. Real-time colour sorting hardware and software integration complicates things. This work uses robotic operating system (ROS) to solve vision-guided colour sorting problems in Cartesian space. Ubuntu 20.04, ROS Noetic, a Raspberry Pi, a camera, and six servos. In Jupyter Lab, unified robotic description format (URDF) is used to build a virtual kinematic model, and Levenberg-Marquardt (LM) optimisation guides object manipulation. OpenCV image processing uses colour conversion, Canny edges, and midpoint estimation to detect coloured objects efficiently. The average servo movement error is 0.46 degrees, and the robot manipulator's final destination positioning error is 1.65 mm. The average object edge detection error is 0.33 mm, and the red, green, blue (RGB) colour distance is 57.84. ROS-based robot manipulator achieves impressive Cartesian space colour sorting accuracy despite image processing challenges, enabling real-world deployment
Optimization model and development of power estimation in photovoltaics for self-load consumption
Get PDFThe Indonesian government's initiative to meet a renewable electricity supply 2050 of around 100% and its ecological benefits, has an impact on the big challenge where the increasing number of photovoltaic (PV) systems due to high power input and risks to electricity network security conditioned sunny weather for a long time. This research aims to provide predictable power output for network operators, thereby enabling longer planning periods and increasing the protection of the national operational network. Independent consumption by household consumers will impact household energy behavior and affect the security of the electricity network. Consumption itself provides an incentive to increase the cost of energy use which will be much higher than the compensation received, so software is needed which can support PV system prediction patterns. This pattern will close the gap in software development in the process of implementing optimization and configuration based on PV estimates. The consumption planning optimization algorithm itself validates the predetermined scenario optimization process. This research contributes to the efficiency and stability of energy use in households