Data Analytics

Massive Open Online Courses (MOOCs) have become increasingly popular since their start in the year 2008. Universities known worldwide for their traditional confined classroom education are also changing their practices by hosting MOOCs. These are Internet-based courses where students can learn at their own pace and follow their own schedule. Study materials and videos are provided that can be used in a blended learning program. Despite its many advantages, it suffers from problems such as high dropout and failure rates. Previous studies have mostly focused on predicting student dropout. This paper contributes to the body of research by investigating both student dropout and result prediction performance of machine learning models built based on different types of attributes such as demographic info, assessment info and interaction with the VLE. An analysis on the OULAD dataset showed that models based on student's interaction with the VLE achieved the high performance in terms of AUC, of up to 0.91 for dropout prediction and 0.93 for result prediction in case of Gradient Boosting Machine.

Studies have identified various risk factors associated with the onset of stroke in an individual. Data mining techniques have been used to predict the occurrence of stroke based on these factors by using patients' medical records. However, there has been limited use of electronic health records to study the inter-dependency of different risk factors of stroke. In this paper, we perform an analysis of patients' electronic health records to identify the impact of risk factors on stroke prediction. We also provide benchmark performance of the state-of-art machine learning algorithms for predicting stroke using electronic health records.

The negative impact of stroke in society has led to concerted efforts to improve the management and diagnosis of stroke. With an increased synergy between technology and medical diagnosis, caregivers create opportunities for better patient management by systematically mining and archiving the patients' medical records. Therefore, it is vital to study the interdependency of these risk factors in patients' health records and understand their relative contribution to stroke prediction. This paper systematically analyzes the various factors in electronic

Parkinson’s disease (PD) is reflected by several motor function disabilities such as tremors, loss of balance, speech impairment, etc. due to loss of dopamine neurotransmitter. While researchers have been building models for diagnosing and classifying PD patients based on gait data, speech data or handwriting data accounting for visible symptoms only, PD patients show non-motor symptoms such as sleep disorder, neuropsychological symptoms, cognitive impairment, olfactory loss, much before the actual diagnosis. This study involves coupling both motor as well as non-motor symptoms of PD patients from up to 10 years longitudinal records in PPMI database and building an optimized deep learning model for PD severity classification based on the Hoehn & Yahr index. This longitudinal complex dataset brings along challenges of dealing with high volume of missing and inconsistent data in various assessments at different time points. To deal with such complexity, the proposed model for this stu...

Attach a completed copy of this sheet to each project (including multiple copies). Attach a Moodle submission receipt of the online project submission, to each project (including multiple copies). You must ensure that you retain a HARD COPY of the project, both for your own reference and in case a project is lost or mislaid. It is not sufficient to keep a copy on computer.

A huge money is involved in pet insurance and medical treatments of pets. Missing pet is a global issue also rescue teams face a lot of challenges while rescuing pets since sometimes they encounter wolfs also which looks same as dogs. For this purpose, new dataset has been created by combining Stanford dog dataset and wolf images from kaggle. Transfer learning technique has been used in this work by using pretrained convolutional networks like ResNet101, ResNet50, VGG16, VGG19, DenseNet201etc. These models can help in distinguishing between wolf and different dog breeds. For evaluating model’s accuracy and model loss metrices has been used. ResNet101 performed better than all other transfer learning techniques and achieved 90.80% accuracy. Many state-of-the-art researches were considered and this work with new dataset of wolf and dog images performed well. One CNN model has also been created which also performed better and further accuracy can be improved by training with more images.

Buying a house is commonly the most important financial transaction for the average person. The fact that most prices are negotiated individually (unlike a stock exchange system) creates an environment that results in an inefficient system. Most people buying houses are inexperienced amateurs with limited information about the market. A housing bubble cannot exist if individuals are making rational decisions. The objective of this paper is to evaluate the performance of a stacked regression model compared to several sub models based on predicting house prices. House characteristics and the final house price was gathered from King County, Washington, USA during the period of May 2014 and May 2015. The observed result indicates that combining the sub algorithms using a general linear model did not significantly improve results.

Lending is one of the main sources of revenue for financial firms and constitutes the main source of risk not just to profitability but to survival. Traditional approaches to credit risk assessment have shown to be flawed and impacted by human biases. The primary goal of this papers is to understand the ability of data mining techniques in their application of identifying potential customers who may default. Our initial findings suggest that data mining can enable a strategy to mitigate risk associated with credit approval.

The prime objective of this research is to identify and exploit a weakness in odds pricing in a pre-selected football betting market. A betting market is usually labelled as being inefficient if there are one or more betting strategies that can generate a profit. Our results indicate that the current market is inefficient and that a sophisticated betting strategy can be profitable over the long run. A widespread misestimation of the home-field advantage was discovered, we successfully exploited this bias for profit gain