|
Predicting Winner of a Professional Basketball Match |
|
Dhananjay Daundkar, Kundan Kandhway(IISER Bhopal, India) |
| Matches in the National Basketball Association (NBA) are followed throughout the world. We use machine learning techniques to predict the winner of an NBA match which has not been played yet. We make use of past performance of the participating teams to make this prediction. Features are computed from the past box scores of the participating teams. Then we use them to train multiple supervised learning classifiers to predict the outcome of a future match. We analyze the model to identify the most important features that affect the predictions the most. Results are presented for 2018 and 2019 regular season matches. Our framework provides a prediction accuracy of about 66%. |
|
|
RDIU-Net: Lightweight Medical Image Segmentation Network |
|
Juon Kurosawa, Armagan Elibol, Nak-Young Chong(JAIST, Japan) |
| we propose a model lightweight medical image segmentation network, Residual Involution U-Net (RDIU-Net). Involution, Residual, and Dense structures are incorporated into the U-Net model to extract both channel and spatial features. Evaluations have been carried out on three different datasets of ultrasound, X-ray, and dermoscopic images. |
|
|
IoT-Based Smart Ankle-Foot Orthosis for Patients with Gait Imbalance |
|
Ferdous Rahman Shefa, Fahim Hossain Sifat, Sayed Chhattan Shah, Muhammad Golam Kibria(University of Liberal Arts Bangladesh (ULAB), Bangladesh) |
| This research aims to transform the conventional Ankle-Foot Orthosis (AFO), a specialized medical brace, into a IoT-based smart device. To facilitate predictive decision-making, the collected data underwent a comprehensive analysis. The analysis was carried out using a variety of machine learning algorithms, providing valuable insights. Among the tested models, the Long Short-Term Memory (LSTM) exhibited the highest accuracy rate, achieving an impressive 95.02% accuracy. This technology-driven approach enhances the monitoring and analysis of AFO usage, enabling personalized treatment plans and optimizing the overall patient experience. |
|
|
Source Component Shift Detection and Classification for Improved Remaining Useful Life Estimation in Alarm-based Predictive Maintenance |
|
Kiavash Fathi(RWTH Aachen University, Germany), Marcin Sadurski(Zurich University of Applied Sciences, Switzerland), Tobias Kleinert(RWTH Aachen University, Germany), Hans Wernher van de Venn(Zurich University of Applied Sciences, Switzerland) |
| This study introduces a solely alarm-based predictive maintenance model for estimating the remaining useful life (RUL) of operational assets. Real-world alarm data from milling machines producing artificial bone joints is employed to develop and test the model. The study emphasizes identifying distinct data modalities (DMs) within alarm data due to varying machine operations and joint sizes. A systematic method for identifying DMs is proposed, enhancing RUL prediction accuracy by up to 25.20% on the test dataset and up to 60.50% for minority DMs. The proposed approach is especially advantageous when limited failure samples exist for specific DMs. |
|
|
Machine Learning Models for Financial Inclusion in Malaysia: Opportunity for Insurance or Takaful in Achieving Financial Inclusion |
|
Hui Shan Lee, Kee Seng Kuang, Ping Xin Liew(Universiti Tunku Abdul Rahman, Malaysia), Annuar Md Nassir(Xiamen University Malaysia, Malaysia), Nazrul Hisyam Ab Razak(Universiti Putra Malaysia, Malaysia) |
| This research harnesses the power of machine learning models to predict insurance uptake in Malaysia. The decision tree model, trained with data, exhibited superior precision and accuracy. Key predictors extracted from the random forest model for insurance adoption were income, employment, and education. Limited documentation stands as the primary obstacle to financial inclusion. The findings suggest there is an opportunity for Takaful players to engage with the government to develop Takaful products to cater to lower-income populations. The application of machine learning techniques within Malaysia's unique financial inclusion landscape constitutes a noteworthy contribution. |
|
|
Deep Learning based EO-LWIR Image Registration Obstacles: Survey |
|
Donyung Kim, Sungho Kim(Yeungnam University, Korea) |
| The field of Electro-Optical and Long-Wave Infrared (EO-LWIR) image registration holds considerable promise in both industrial and personal device applications. Despite its potential, successful EO-LWIR registration has remained elusive due to the fundamental differences between the EO and LWIR sensors. EO sensors primarily use reflected light to capture images, while IR sensors utilize radiant light reflected and emitted from object source. This divergence gives rise to three key obstacles. In this paper, we conduct a comprehensive analysis of various EO-IR methods, evaluating their capacity to address these specific problems. |
|