22505 research outputs found
Sort by
Achilles
No full textACHILLES is a speculative design project and conceptual artifact. It is not intended to function as a medically certified or biomechanically operational prosthetic arm. Instead, it operates within the realm of design research, using visual realism and luxury aesthetics to explore how assistive technologies could be perceived differently in cultural, emotional, and commercial contexts. Using 3D software to experiment with color, pattern, and texture, I designed a realistic mockup to demonstrate how Achilles could appear in a high-end commercial setting. Instead of animating the model, I focused on camera motion to highlight the prosthetic from multiple angles, emphasizing its sculptural and ornamental qualities. This project addresses a topic that is often difficult to discuss and reframes it through design, bringing awareness to the experience with elegance, empathy, and dignity. Rather than avoiding the conversation around limb loss, Achilles invites people to engage with it in a more uplifting, empowering way
Cosmic Legends
No full textCosmic Legends is a touchscreen-based interactive kiosk that invites users to explore how different cultures interpret the same star—Betelgeuse. Designed for public educational settings, it blends visual storytelling, mythology, and voice interaction into a cohesive learning experience. This project challenges the Eurocentric approach of most astronomy tools by introducing global narratives rooted in Navajo, Tupi, and Japanese sky traditions. Users begin by tapping a glowing version of Betelgeuse, which transforms the night sky into a new cultural context. Each sky shift is paired with narrated myths, synchronized captions, and celestial motion design. The interface prioritizes accessibility with dyslexia-friendly text, high-contrast visuals, and simplified navigation for large-format kiosks. The thesis explores how interaction design can bridge science and story, supporting a more inclusive understanding of the stars. With speculative layers such as AI-curated myths and AR overlays, Cosmic Legends offers a framework for designing empathetic, narrative-driven educational tools
The Klik Chair - Designed to Move, Built to last
No full textFor the modern individual who moves housing frequently, buying furniture has become an increasingly tedious and frustrating process. Traditional furniture often presents challenges due to its bulk, weight, and complex assembly requirements, making it difficult to transport and reassemble when moving to a new living space. As a result, the true value of a piece of furniture is no longer determined solely by its price tag or aesthetic appeal but by its ability to adapt to a nomadic lifestyle. Additionally, furniture that holds long-term sentimental or functional value has become even more significant in an era where disposability is routine. By prioritizing durability and timeless design, furniture can transform from a temporary necessity into a cherished possession that moves with its owner, rather than being left behind or discarded. A chair that has a simple and toolless assembly process, is proposed to address these challenges, offering a design that retains both functional and emotional value over multiple relocations. A well-designed piece must be easy to assemble and disassemble without specialized tools, ensuring it can seamlessly transition from one home to another. The “Klik” chair design provides users with a relaxing seating option that is sturdy, simple, and aesthetically pleasing. With tool-less assembly, customizability, and repairability, the chair becomes a functionally relevant part of the user’s life
Optimizing Human Resource Decisions: Predicting Promotions Using Data Analytics
No full textProper and objective selection of high potential employees to promote them is a major dilemma in the Human Resources (HR) department, more so in sensitive and hierarchal environments in the public sector where subjectivity is likely to take place. This paper is based on this ubiquitous issue, and it seeks to develop, experiment, and examine a clear and equitable machine learning model that can forecast the possibility of an employee to get a promotion according to organized past HR records. The technique was solid preprocessing, alleviation of extreme class imbalances on the basis of the Synthetic Minority Over-sampling Technique (SMOTE), as well as comparative examination of intricate (XGBoost) and very clarifiable (Logistic Regression) classifiers. However, the XGBoost model has the highest predictive power (F1-Score: 0.4909), but the Optimized Logistic Regression model has an alternative F1-Score of 0.4639; this is why the latter is chosen in favor of the former because it is the most critical in terms of transparency and auditability in the public sector. The important features analysis revealed that organizational structure (departmental membership), the long term past performance, and tenure played a key role in promotion. Notably, a fairness test showed that there was predictive equity in terms of genders, which proved the ethical value of the model. The research presents a valid, objective and gender-neutral decision support framework, which offers a needed entry point via which establishments can pay more attention to meritocracy and confidence in their talent handling processes by individuals
Deepfake Audio Detection
No full textThe rise of deepfake audio technology has introduced a serious threat to information credibility, personal security, and media integrity. This thesis investigates the application of machine learning techniques for detecting synthetic audio through the analysis of acoustic features, including Mel-Frequency Cepstral Coefficients (MFCCs), spectral centroid, chroma_stft, and zero-crossing rate. The dataset used in this study was sourced from Kaggle and contains labeled samples of real and fake audio clips. The research aimed to train and evaluate multiple machine learning models—Support Vector Machines (SVM), Random Forest, XGBoost, Logistic Regression, and Neural Networks—to determine the most effective approach for deepfake audio classification. Each model was assessed using standard performance metrics such as accuracy, precision, recall, F1-score, and ROC-AUC. The study followed the CRISP-DM methodology, encompassing data preprocessing, feature extraction, model training, and performance evaluation to ensure methodological rigor and reproducibility. The findings reveal that the Neural Network model consistently outperformed all other classifiers, achieving 98% accuracy when trained on all variables and maintaining 96% accuracy, recall, and precision when using ten features selected through Random Forest. This demonstrates the model’s robustness, efficiency, and capacity for generalization, even with a reduced feature set. In contrast, traditional models such as Logistic Regression and LSVM achieved accuracies around 92%, while their performance decreased notably after dimensionality reduction. Sensitivity analysis and partial dependence plots further confirmed that key features—particularly rms and MFCC components—had the strongest influence on classification outcomes. Overall, this research demonstrates that combining deep learning with optimized acoustic feature selection enables accurate and interpretable detection of deepfake audio. The study contributes a scalable and generalizable detection framework that can be applied to real-world verification systems, supporting advancements in digital forensics, cybersecurity, and media authenticity
Optimizing Waste Management and Recycling Patterns Using Data Analytics
No full textThe increasing pace of urbanisation and consumption has increased the burden of waste generation in the world and it is a huge burden on the current waste management systems. The United Arab Emirates (UAE) is a region that is intensifying this challenge through the accelerated urbanization, high sustainability targets, and the necessity of effective recycling policies. The thesis that is being examined explores the ways in which the data analytics can be utilized to streamline waste management and recycling trends, emphasizing the enhancement of the collection process, forecasting the trends in the waste generation as well as facilitating the use of the data in making the decisions within the area of the municipality. This research is based on the CRISP-DM (Cross-Industry Standard Process of Data Mining) model which offers a methodical way of converting unprocessed municipal information into actionable information. Based on a synthesis of municipal waste collection reports, IoT-enabled smart bin sensors data, and Geographic Information System (GIS) data, the study implements both quantitative and spatial analytics in recognizing inefficiencies, the high-waste areas and the trends in recycling behaviors in the chosen urban districts within the UAE. The six main questions that guided the research were (1) in what way data integration with multi-source information can increase predictive accuracy, (2) what spatial and behaviour factors can affect the performance of recycling, (3) howmachine learning models could be used to optimize the collection routes, (4) how the CRISP-DM can be relevant to municipal waste analytics, (5) how the use of data-driven dashboards can benefit operational decisions, and (6)what practical benefits are possible through predictive scheduling. Python (Pandas, Scikit-learn, XGBoost) and RStudio were used to process and model the data and supported by QGIS to visualize the data spatially. The models showed that predictive analytics can increase the efficiency of routes by 25-30% and decrease the events of overflow by 20 which are consistent with other international studies of the same nature. Clustering analysis showed that population density, land use type, and recycling performance have significant correlation and time series forecasting can correctly represent the periods of highest waste generation. The results show that combining the information collected by IoT sensors with the municipal data can help municipalities transition to proactive waste management. A decision dashboard written in Streamlit was also built to mimic the real-time route optimization, overflow warning, and recycling rate, which showed how analytics could directly guide operational planning. Results of this study support the conclusion that data-driven waste management can play an important role in supporting the objectives of the Zero Waste 2030 of the UAE by minimizing operating expenses, improving sustainability, and increasing the satisfaction of citizens. Some proposals include the necessity of unified municipal data system, constant investment into IoT infrastructure, and incorporation of data on citizen engagement in future designs. Further studies must consider the future opportunities of applying deep learning to automated image-based waste classification and predictive frameworks in a rural and semi-urban setting. These attempts would also improve the smart, circular, and sustainable urban development of the UAE
Optimizing Delivery Time Predictions Using Machine Learning: A Data-driven approach to Last-mile Logistics
No full textThis paper addresses how machine learning can be utilized to improve the prediction of delivery times during the last mile, specifically in Dubai urban logistics issues whereby the traffic congestion and weather circumstances usually contribute to unpredictable delivery times. The main goal was to evolve machine learning models that can predict properly delivery time depending on several parameters, i.e., speed of traffic, weather, length of delivery, and geography. The reason why three machine learning models were chosen [Random Forest, Gradient Boosting, and XGBoost] in this analysis is that they have the opportunity to work with non-linear relationships in the data. Models were trained and assessed based on the Metrics of the Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and the R-Squared (R 2 ). The findings indicated that Grade Boosting was more effective in comparison with the other models through the highest R2 of 0.6384. Geospatial analysis was also used in the research to examine how the location of delivery affected the delivery time. Through the geocoding of delivery points and mapping of their presence, which exposed the geographic regions in which the delivery was comparatively longer due to various factors, including, but not limited to traffic issues and weather pattern. The results show that the efficiency of last-mile delivery operations can be substantially increased by the real-time data integration, better feature engineering, and spatial optimization. The research adds to the body of knowledge focusing on data-based logistics and offers practical recommendations that the logistics firms could use to improve their finetuning delivery time estimations and efficiency of their operations
Call for Papers
Get PDFYou are invited to submit a paper for possible inclusion in the first issue of the Journal of Peace, Conflict, and Security Studies (JPCSS). JPCSS is housed in the Political Science and Public Policy department of the Rochester Institute of Technology-Kosova (RITK). This is an international peer-reviewed journal published as a joined venture of the RIT main campus in Rochester, NY, RITK in Prishtina, and the Kosovo Defence Academy
Resilient Power System Load Frequency Control
Get PDFStandard frequency and voltage levels need to be maintained when the electric power system is running in order to keep it in a satisfactory state of operation. Between AC power generation and use, there should be a balance between active and reactive power. However, due to variations in active and reactive power requirements, the frequency and voltage provided may fluctuate from their rated values. Any unwelcome change in frequency or voltage can have an immediate effect on how the power system operates and may even cause harm to linked devices. These parts are intended to operate at certain rated frequencies and voltage levels, and any variation could trip connected loads and power producing equipment. For maintaining a constant frequency, an effective load frequency control (LFC) design is required. A proportional-integral-derivative (PID) controller is used by a load frequency control (LFC) to keep the frequency near the rated values. By adjusting its proportional, integral, and derivative gains, the PID controller lowers the absolute errors of the frequency deviation that make up its integral. This study suggests two techniques for adjusting the PID controller parameters: tuning for artificial neural networks and Particle Swarm Optimization (PSO). On a single-area system, both tuning techniques are used in the LFC loop, and their performance is contrasted in terms of overshoot, undershoot, and settling time. The suggested tuning strategies outperform more traditional methods in terms of transient stability and steady-state stability