Computational Sciences - Master's theses

Permanent URI for this collection

https://laurentian.scholaris.ca/handle/10219/2096

Browse

Now showing 1 - 20 of 104

The problematic internet use in Pakistan
(2021-06-25) Hassan, Hammad
The revolution of smart devices, and smart objects has dramatically improved the usability of the Internet around the world. It’s happening, and over the last few decades, we’ve seen such a dynamic the Internet trend. The increasing use of the Internet today has played a very significant role in the compulsion of the Internet. It can affect the educational, psychological, medical, and social well- being of the user. The Internet restriction in developing countries like Pakistan are becoming more severe, and the public is not fully aware of the Internet usage. The current COVID-19 and subsequent lockdowns situation have further raised the level of the Internet coercion in developing countries like Pakistan. This treatise explores existing literature, social dilemmas, and problematic Internet use in Pakistan. We look forward to formally analyzing the literature and conducting pilot studies to make further contributions to this issue.
Outcome-based judgement categorization of the Supreme Court of Canada
(2022-09-12) Malley, Thomas
Outcome-based judgement categorization of the Supreme Court of Canada (SCC) focuses on the multidisciplinary field of computational law. Regarding court hierarchy, the SCC is the highest court in Canada. Decisions from this court generally bind any lower court. Since court decisions are in a textual format, it is possible to correctly categorize outcomes of the SCC utilizing Natural Language Processing (NLP) techniques. The experiment contained shows algorithmic categorization performance F1 greater than 60. This result is significant given the binary nature of case outcomes (allow, dismiss) that an individual unfamiliar with the law should be able to guess 50% of the time correctly. This work is a preliminary study of future work to indicate the possibility of outcome forecasting in the judicial branch of the government.
Evaluation of U-Net model in the detection of cervical spine fractures
(2023-10-18) Kheirandish, Faranak
The cervical spine is composed of seven vertebrae from C1 to C7 with a lordotic curve (C-shaped curve) and joints between vertebrae for spine mobility. A computed tomography (CT) is commonly used by experts and physicians in imaging diagnosis to give information about the cervical spine and vertebrae in the neck. Diseases such as spinal stenosis (narrowing of the spinal canal), herniated discs, tumors, and fractures in the cervical spine can be diagnosed by CT scans. Quickly detecting the presence, and location of cervical spine fractures in CT scans helps physicians prevent neurologic deterioration and paralysis after trauma. Throughout this thesis, a U-Net model was trained for semantic segmentation on approximately 2019 study instances with provided CT images, while only 87 of them have been segmented by spine radiology specialists. After that, a combination of 2D CNN and bidirectional GRU deep learning models was used for the detection of fractures in each vertebra, as a classification task. The objectives of this research are to develop two deep-learning models for detecting and localizing cervical spine fractures and evaluate the ongoing research activities on semantic segmentation and classification in the medical field. This research aims to use a semantic segmentation algorithm in deep learning by using U-Net architecture to estimate the location of each cervical vertebra, as well as propose a deep convolutional neural network (DCNN) with a bidirectional GRU memory (Bi-GRU) layer for the automated detection of cervical spine fractures in CT images. This approach was trained and tested on a dataset provided by RSNA (a team of the American Society of Neuroradiology and Spine Radiology). Furthermore, the critical factors, such as preprocessing techniques and specialized loss functions were explored that must be taken into consideration when segmenting 3D medical images. Whether used as a standalone framework for segmentation and classification tasks or as an integrated backbone for medical image processing, this architecture is flexible enough to accommodate other models. The proposed approach yields results that are comparable to those of existing techniques, but it can be improved by using larger image sizes and more advanced GPU workstations that will reduce the overall processing time. Future research will be using other pretrained networks as an encoder and increasing image sizes to examin the performance improvmet of the architecture which needed more advanced computational resources and also integrate the current architecture into a simulated crash scenarios to use in various applications such as producing protecive sport equipments.
Emotion-centric image captioning using a self-critical mean teacher learning approach
(2022-11-07) Yousefi, Aryan
Image Captioning is the multi-modal task of automatically generating natural language descriptions based on a visual input using various Deep Learning techniques. This research area is in the intersection of Computer Vision and Natural Language Processing fields, and it has gained an increasing popularity over the past few years. Image Captioning is an important part of scene understanding with various extensive applications, such as helping visually impaired people, recommendations in editing applications, and usage in virtual assistants. However, most of the previous work in this topic has been focused on purely objective content-based descriptions of the image scenes. The goal of this thesis is to generate more engaging captions by leveraging humanlike emotional responses in the captioning process. To achieve this task, a Mean Teacher Learningbased method has been applied on the recently introduced ArtEmis dataset. This method includes a self distillation relationship between the memory-augmented language models with meshed connectivity, which will be first trained in a cross-entropy based phase, and then fine-tuned in a Self-Critical Sequence Training phase. In addition, we propose a novel classification module by decreasing texture bias and encouraging the model towards a shape-based classification. We also propose a method to utilize extra emotional supervision signals in the caption generation process, leveraging the image-to-emotion classifier. Comparing with the state-of-the-art results on ArtEmis dataset, our proposed model outperforms the current benchmark significantly in multiple popular evaluation metrics, such as BLEU, METEOR, ROUGE-L, and CIDEr
Biocybernetic closed-loop system to improve engagement in video games using electroencephalography
(2022-01-06) Klaassen, Stefan
The purpose of this paper was to determine the level of engagement with a specific stimuli while playing video games. The modern video game industry has a large and wide audience and is therefore becoming more popular and accessible to the public. The interactions and rewards offered in video games are a key to keep player engagement high. Understanding the player’s brain and how it reacts to different type of stimuli would help to continue improving games and advance the industry into a new era. Although studying human engagement had started many years ago, the application of measuring it in video game players has only been applied more recently and is still an evolving field of research. This thesis will be taking an objective approach by measuring engagement through electroencephalogram (EEG) readings and seeing if it will help improve current dynamic difficulty adjustment (DDA) systems for video games leading to more engaging and entertaining games. Although statistically significant findings were not found in this experiment, the technique for future experiments were laid out in the form of classifiers comparison and program layouts.
Crop disease detection using deep learning techniques on images
(2023-05-23) Deputy, Kinjal Vijaybhai
Agriculture is a field which is referred to as the main sector for the development of the economy in various countries, and it is also providing food to the large population of the world despite various limitations and boundaries. Food security is threatened by several factors including climate change, the decline in pollinators, plant diseases and others. Different efforts have been developed to prevent crop loss due to infections in the plants. The advancement in technology is helping farmers in developing different systems that can help in reducing the problem. Smartphones specifically offer very novel ways to identify diseases because of their computing power, high resolution displays, and extensive built-in sets of accessories, such as advanced HD cameras. This leads to a situation where disease diagnosis based on automated image recognition is needed. Image recognition is made possible by applying a deep learning approach. So the research is aimed to analyze deep learning-based image detection techniques to identify the various diseases in the plants. The “PlantVillage” dataset has been used to train models. Deep learning Architectures such as AlexNet and GoogleNet, ResNet50 and InceptionV3 are used. Two approaches are used to train the model: ‘training from scratch’ and ‘transfer learning’. It was found from the results of the primary analysis that the GoogleNet leaves behind the AlexNet, ResNet50 and InceptionV3 in training from scratch approach. And ResNet50 performed best in transfer learning.
Optimal data allocation method considering privacy enhancement using E-CARGO
(2023-04-19) Peng, Chengyu
With the rise in popularity of cloud computing, there is a growing trend toward the storage of data in a cloud environment. However, there is a significant increase in the risk of privacy information leakage, and users could face serious challenges as a result of data leakage. In this paper, we propose an allocation scheme for the storage of data in a collaborative edge-cloud environment, with a focus on enhanced data privacy. In addition, we explore an extended application of the approach to sourcing. Specifically, we first evaluate the datasets and servers. We then introduce several constraints and use the Environments-Classes, Agents, Roles, Groups, and Objects (E-CARGO) model to formalize the problem. Based on the qualification value, we can find the optimal allocation using the IBM ILOG CPLEX Optimization (CPLEX) Package. At a given scale, the allocation scheme scores based on our method improve by about 50% compared to the baseline method and the trust-based method. Moreover, we use a similar approach to analyze procurement issues in the supply chain to help companies reduce the carbon emissions. This shows that our proposed solution can store data in servers that better suit their requirements and is adaptable to other problems.
Detecting span in emails using advanced machine learning methods
(2022-06-03) Mistry, Nirali
E-mail is one of the quickest and most professional ways to send messages from one location to another around the world; however, increased use of e-mail has increased to received messages in the mailbox, where the recipient receives a large number of messages, some of which cause significant and varied problems, such as the theft of the recipient's identity, the loss of vital information, and network damage. These communications are so harmful that the user has no way of avoiding them, especially when they come in a variety of forms, such as adverts and other types of messages. Spam is the term for these emails Filtering is used to delete these spam communications and prevent them from being viewed. This research intends to improve e-mail spam filtering by proposing a single objective evaluation algorithm issue that uses Deep Learning, Genetic Algorithms, and Bayes theorem-based classifiers to build the optimal model for accurately categorizing e-mail messages. Text cleaning and feature selection are used as the initial stage in the modeling process to minimize the dimension of sparse text features obtained from spam and ham communications. The feature selection is used to choose the best features, and the third stage is to identify spam using a Genetic algorithm classifier, Support Vector Machine, Bayesian Classifier, Nave Bayes, SVM, Random Forest, and Long-Short Term Memory classifier.
BERT-based multi-task learning for aspect-based sentiment analysis
(2022-01-20) Bhagat, Yesha
The Aspect Based Sentiment Analysis (ABSA) systems aims to extract the aspect terms (e.g., pizza, staff member), Opinion terms (e.g., good, delicious), and their polarities (e.g., Positive, Negative, and Neutral), which can help the customers and companies to identify product weaknesses. By solving these product weaknesses, companies can enhance customer satisfaction, increase sales, and boost revenues. There are several approaches to perform the ABSA tasks, such as classification, clustering, and association rule mining. In this research we have used a neural network-based classification approach. The most prominent neural network-based methods to perform ABSA tasks include BERT-based approaches, such as BERT-PT and BAT. These approaches build separate models to complete each ABSA subtasks, such as aspect term extraction (e.g., pizza, staff member) and aspect sentiment classification. Furthermore, both approaches use different training algorithms, such as Post-Training and Adversarial Training. Moreover, they do not consider the subtask of Opinion Term Extraction. This thesis proposes a new system for ABSA, called BERT-ABSA, which uses MultiTask Learning (MTL) approach and differentiates from these previous approaches by solving all three tasks such as aspect terms, opinion terms extraction, and aspect term related sentiment detection simultaneously by taking advantage of similarities between tasks and enhancing the model’s accuracy as well as reduce the training time. To evaluate our model’s performance, we have used the SemEval-14 task 4 restaurant datasets. Our model outperforms previous models in several ABOM tasks, and the experimental results support its validity
Pairwise comparisons and visual perceptions of 3D shape volume estimation
(2022-05-31) Wan, Wenjun
Using pairwise comparisons for estimations increases accuracy. At present, scholars use the pairwise comparisons method to make subjective comparison between one-dimensional image and two-dimensional images. This research is about the subjective comparison of three-dimensional images. We rst sets a xed object volume and then uses the random method to generate multiple three-dimensional objects with di erent shapes and then scale them to our designed volume values. This study also virtualizes and binarizes the image and prints the actual object in the way of 3D printing for respondents to observe. Thirty-two respondents used the direct and pairwise comparisons methods to rate the volume of ve randomly generated 3D shapes. It is found that using the direct method, the observer's estimation errors is higher (in average) than when the paired comparisons method is used. The pairwise comparisons method can improve the accuracy of estimating the volume of random objects.
Social media hate speech detection using explainable AI
(2022-05-25) Mehta, Harshkumar
Artificial Intelligence has invaded various fields in the present times. Be it science, education, finance, business or social media, Artificial Intelligence has found its applications everywhere. But currently, AI is limited to only its subset ‘Machine Learning’ and has not even realized its full potential. In machine learning, in contrast to traditional programming which requires writing algorithms, it is required to find the algorithm that learns patterns from a given dataset and builds a predictive model and the computer learns the patterns between input and output based on that. However, a key impediment of current AI-based systems is that they often lack transparency. The current AI systems have adopted a black box nature which allows powerful predictions, but these predictions cannot be explained directly. To gain human trust and increase transparency of AIbased systems, many researchers think that Explainable AI is the way forward. In today’s era, an enormous part of human communication takes place over digital platforms, for example, through social media platforms and so does hate speech, which are dangerous for an individual person as well as the society. These days automated hate speech detection is built on social media platforms such as Twitter, Facebook, etc. using machine learning approaches. Deep learning models attain a high performance has low transparency due to complex models, which leads to “trade-off” between performance and explainability. Explainable Artificial Intelligence (XAI) was used to create black box approaches interpretable, without giving up on performance. These XAI methods provide explanations that can be translated by humans without having a depth of knowledge in deep learning models. XAI characteristics have flexible and multifaceted potential in the hate speech detection by the deep learning models. XAI thus provides a strong interconnection between an individual moderator and hate speech detection framework, which is a pivot for the research study in interactive machine learning. In the case of Twitter, the main tweets are detected for hate speech however retweets and replies are not detected for hate speech as there is no tool to handle the task to detect the hate speech for in progress conversations. Interpreting and explaining decisions made by complex AI models to understand the decision-making process of these model is the aim of this research. While machine learning models are being developed to detect the hate speech on social media, these models lack the interpretability and transparency on the decisions made. Traditional machine learning models achieve high performance at the cost of interpretability and explaining model decisions. The main objectives of this research are, to review and present a comparison of various techniques used in Explainable Artificial Intelligence (XAI), to present a novel approach for hate speech classification using Explainable Artificial Intelligence (XAI) and, to achieve a good trade-off between precision and recall for the method proposed. Explainable AI models for hate speech detection will help social media moderators and any other users for these models to not only see but also study and understand how the decisions are made and how the inputs are mapped to the output. As a part of this research study, two data sets were taken to demonstrate Hate Speech Detection using Explainable Artificial Intelligence (XAI). Data preprocessing was performed to remove any bias, clean data of any inconsistencies, clean the text of the tweets, tokenize, and lemmatize the text, etc. Categorical variables were also simplified in order to generate a clean dataset for training purposes. Exploratory data analysis was performed on the data sets to uncover various patterns and insights. Various pre-existent models were applied to the Google Jigsaw dataset such as Decision Trees, K-Nearest Neighbours, Multinomial Naïve Bayes, Random Forest, Logistic Regression, and Long Short-Term Memory (LSTM) out of which LSTM achieved an accuracy of 97.6%, which is an improvement compared to the studies of Risch et al. (2020). Explainable method like LIME (Local Interpretable Model-Agnostic Explanations) is applied on HateXplain dataset. Variants of BERT (Bidirectional Encoder Representations from Transformers) model like BERT + ANN (Artificial Neural Networks) and BERT + MLP (Multilayer Perceptron) were created to achieve a good performance in terms of explainability using ERASER (Evaluating Rationales and Simple English Reasoning) benchmark by DeYoung et al. (2019) where in BERT + ANN achieved better performance in terms of explainability as compared to the study by Mathew et al. (2020).
Analyzing impact on bitcoin prices through Twitter social media sentiments
(2022-04-28) Patel, Jay
Many cryptocurrencies exist in today's date, and many more are on the verge of being brought into circulation. It is a form of a digital currency but instead of being run by a centralized authority and government, it is a decentralized structure that is created using blockchain technology. These currencies are highly influential and unpredictable with their factors of influence ranging high and low all over the world. This research revolves around the most well-renowned cryptocurrency which is Bitcoin. The focus here is on the discussion around the relationship of bitcoin with the prominent online media platform called Twitter. Twitter has been taking part in the discussion of almost all major as well as related incidents and events all around the world. It is a social media platform that is informative as well as useful for the public so much, that even major personalities, as well as politicians, take to the platform in order to express their views quickly on an important matter. The research included firstly gathering the tweets and was divided into two parts - Verified and Non-Verified users and then a cleaning process was done on the data to make sure that only the desired and necessary data if left for further research. The tweets regarding bitcoin were analyzed and utilized for a deeper observation so that the sentiment can be extracted and can be visualized against the bitcoin prices to derive a conclusion regarding the relationship between Twitter and Bitcoin prices. The analysis returned a lot of insights as well as inference relating to the influence that the Bitcoin prices and related tweets have on each other. The results of the report mention the outcome of the analysis that was found stating the original hypothesis to be true or not
Non-directional forecasting of stocks using Twitter’s quantitative metrics
(2022-04-27) Keshmiri, Vahid
Using public opinion to forecast political, social, and financial phenomena has usually been an attractive subject for researchers. With the unprecedented growth of social networks, along with the growth of technology and data mining methods, we have seen researchers move to platforms such as Twitter to predict political, social, and economic indicators. However, most researchers have turned to sentiment analysis of posts on social media, and mostly due to the quality of the output of sentiment analysis algorithms, have reached results with not good accuracy. This thesis seeks to prove the assumption of a significant relationship between quantitative data extracted from social networks and economic indicators especially focused on stock value changes of certain brands and companies. The results presented in this thesis are based on determining the severity or extent of the change in stock value, regardless of the direction of change. In other words, the presented outcome is for non-directional prediction. Due to time and financial constraints, this thesis focuses on proving the existence of a significant relationship between stock price changes of three brands, Apple, IBM, and Pfizer, and measurable Twitter indicators containing likes, replies, followers, quotes, and retweets. To find out how feasible is this method, four machine learning algorithms are applied, namely Artificial Neural Network, Random Forest, Support Vector Machine, and Adaptive Boost. Except for SVM, the results are satisfying and promising, especially with AdaBoost.
Monte Carlo simulated heat transport in semiconductor nanostructures
(2022-10-31) Gibson, Graham
Nanoscale energy transport is a topic of considerable interest as heat transport at these scales can no longer be accurately predicted by diffusion theory. An alternative approach is to use the Boltzmann transport equation, but this equation is challenging to solve in the case of phonon transport, and its exact resolution is currently one of the open research subjects in mathematics. A program has been developed to study nanoscale heat transport by solving the Boltzmann transport equation using two variations of a phonon Monte-Carlo method. The first variation is primarily derived from the works of Mazumber and Majumber (2001). The second variation follows the process of Peraud and Hadjiconstantinou (2012). While both variations follow methodology from existing works, the implementation details are unique. The simulation procedures differ from existing methods by incorporating a ‘system evolution’ algorithm that allows temperatures throughout the system to be periodically updated while simulating phonons one-by-one. The resulting software can rapidly simulate heat transport in relatively complex geometries. The Monte Carlo portion of the software is implemented using parallelized C++ code. Simulating phonons one-by-one makes the parallelization scheme natural and straightforward, although more sophisticated parallelization schemes may result in further computational speedup. The user input is a self-documenting JSON file generated via a Python script. The software is used to study thermal transport through various silicon and germanium nanostructures. Benchmark simulation testing shows that the temperature profiles produced by the simulations largely agree with analytical results and results from the literature, as does the predicted thermal conductivity. However, the thermal conductivity is quite sensitive to the relaxation rates that are used. While both variations of the phonon Monte Carlo method presented in this study strike a good balance between accuracy and efficiency and retain an intuitive connection to the problem physics, a noticeable difference in computational efficiency and precision is observed. With the exceptions of low-temperature ranges and possibly systems with extreme temperature differences, the second variation should be preferred when considering computational performance and precision.
An analysis of lung cancer survival using multi-omics neural networks
(2022-01-27) Naik, Krinakumari
A key goal of precision health medicine is to improve cancer prognosis. Despite the fact that numerous models can forecast differential survival from data, progressive algorithms that can assemble and select important predictors from progressively complex data inputs are urgently required. As a result, these models should be capable to provide more information about which types of data are most significant for improving prediction. Because they are adaptable and account for data density in a non-linear manner, deep learning-based neural networks may be a feasible solution for both difficulties. In this study, we use Deep Learning-based networks to get how gene expression data predicts Cox regression survival in lung cancer. SALMON (Survival Analysis Learning with Multi-Omics Neural Networks) is an algorithm that collects and simplifies gene expression data and cancer biomarkers in order to enable prognosis prediction. When more omics data was comprised in model construction, the results (concordance index = 0.635 and log-rank test p-value = 0.00881) showed that performance improved. We employ eigengene modules from the results of gene co-expression network analysis as model inputs in its place of raw gene expression principles. This algorithm verified specific mRNA-seq co-expression modules and clinical information, which show crucial roles in lung cancer prognosis, revealing various biological functions by exploring how each contributes to the hazard ratio. SALMON also performed well compared to other Deep Learning Survival prognosis models.
Generalized temperature-driven insect population dynamics model – a mechanistic approach
(2023-01-18) Delay, Dominique
Demand for computer models that simulate insect population dynamics is growing due to many factors including increased pressure on natural resources and climate change. Generalized models are a practical way to simulate multiple insect species with a single computer model, reducing the time spent developing species-specific models for each insect of interest. In this thesis, a generalized insect population dynamics model is presented. The model uses a mechanistic approach, leveraging data on underlying population drivers such as temperature-dependent vital rates to simulate changes in a population. The general model structure and code were adapted from the species-specific Drosophila suzukii model by Langille et al. (2016). The species-specific model was modified to account for a variety of insect species, minimise the number of required parameters, and use parameters that are available through literature, ensuring the general model’s simplicity and ease of use. Through exploration and sensitivity tests, the model’s elements were found to largely behave as expected from the real-life systems. The model was also validated for its intended use as a non-predictive, exploratory model through the comparison of published field or simulation population studies. The model successfully approximated published population studies when simulating insect species with simple life cycles, however, simulations of insect species with more complex life cycles, or social structures, were not as successful. Overall, despite some limitations, the general model presented in this thesis can simulate many insect species population dynamics and is ideal for study ideation, prototyping, and rapid exploration.
Fine-tuning a general transformer model on story-lines of IMDB movies database
(2022-01-13) Ghasemi, Hojat
Recent transformer-based language models pre-trained on huge text corpora have shown great success in performing downstream Natural Language Processing (NLP) tasks such as text summarization when fine-tuned on smaller labeled datasets. However, the impact of fine-tuning on improving the performance of pre-trained language models in summarizing movie storylines have not been explored. Moreover, there is a lack of extensive labelled datasets containing movies storylines to allow pre-trained language models delving deeper in this realm. In this research work we propose a novel labelled dataset containing IMDB movie storylines alongside their summaries for teaching pre-trained language models how to perform text summarization on movie storylines. Furthermore, we showcase the potential of this dataset by fine-tuning a T5-base model with the use of this dataset. Our results show that fine-tuning a T5-base model on this dataset can significantly improve the performance in summarizing movie storylines
Inflation forecasting using hybrid ARIMA-LSTM model
(2022-01-07) Jamil, Hira
Prediction of time series is one of the most demanding research areas due to the nature of various time series i.e., stocks, inflation, stock indexes etc. Various methods have been used in the past to forecast such time series, however, Machine Learning (ML) methods have been suggested in the academic literature as alternatives to statistical ones for time series forecasting. Yet, scant evidences are available about their relative performance in order of their accuracies and computational requirements. In this thesis, a hybrid model consisting of ARIMA and LSTM is proposed and compared with individual models ARIMA, LSTM, and PROPHET for inflation forecasting. Two Scale-dependent metrics namely mean absolute error (MAE) and root mean square error (RMSE), one Percentage-error metric, mean absolute percentage error (MAPE) and coefficient of determination (R 2 ) are used to evaluate the variance between dependent and independent parameters for inflation forecasting in developed and developing countries. Consumer Price Index (CPI) data is collected monthly to reflect the effect of price inflation at consumer level. Most of the central banks depend on inflation forecast to inform their respective monetary policy makers and to enhance the efficacy of monetary policy. The publicly available CPI data is presented for analysis and evaluation of price inflation effects on developed and developing countries. For this research work, six developed countries (Canada, United States, Australia, Norway, Poland and Switzerland) and six developing countries (Colombia, Indonesia, Brazil, South Africa, India and Mexico) with different durations are targeted to evaluate the performances of proposed machine learning model and the individual models to forecast inflation (CPI). The proposed HYBRID model with one-step ahead forecasting outperformed every other model for forecasting inflation (CPI) of developed and developing countries regardless of duration. The best performance was observed by taking 90% training data and 10% testing data. All iv forecasting models performed better on data of six developed countries with overall average errors of 1.023796 in MAE, 0.009648 in MAPE and 1.222454 in RMSE when taking 10% as test data. While in the case of developing countries overall average errors of MAE, MAPE and RMSE was 1.361308, 0.011847, and 1.562288 respectively. Also, in the case of 20% and 30% test data, the performance of all models on developed countries data was better than developing countries in terms of least errors in MAE, MAPE and RMSE.
Gender human-computer interaction: investigating the perceived credibility of mobile applications from gendered perspective
(2021-11-04) Randhawa, Gulnaz Kaur
Consumer market has witnessed remarkable shift from designer’s preferences to what actual users want. Not only has this led to hike in market profit but also resulted in products with enhanced User-Experience. In the field of HCI, designers have been working to make their end products usable to large group of users belonging to different ages, gender as well as techsavvy and non-computer literates. Considering ‘gender’, much research has been done and is still being conducted to explore its’ relevance to Human-Computer Interaction. This research addresses the researcher’s vision to identify gender differences in User-Experience received from using mobile applications. The idea isto experimentally prove whether there exists samegender or opposite gender credibility in rating the usability of mobile applications. If no statistically significant differences are noted in male and female respondents’ respective ratings of applications from male and female designer, then null hypothesis will be accepted. This would imply that UI designers did not target users of any particular gender during the design phase thereby producing gender-inclusive designs. On the contrary, if differences are found then alternate hypothesis shall be accepted which will be an indication of gender bias being propagated in designs. Data collected from 30 participants (15 males and 15 females) showed no statistically significant differences in the ratings of male and female designer’s interfaces. Future research has been proposed with greater sample size along with additional test variants such as virtual gendered avatars.
Prediction of solar radiation values on Indian cities using machine learning and deep learning techniques
(2021-10-28) Patel, Meshwa
The use of renewable resources has grown rapidly as non - renewable resources become exhausted, electricity consumption rises, and environmental problems develop. Rapid improvements in energy sources have opened possibilities for using solar energy to generate electricity. Solar panels were used for various purposes, including sun heating, dining, and building lighting. These regular long estimates of global solar radiation from a specific place were critical for effective solar power generation and utilization. The climate patterns of a region determine the quantity of sunlight radiation that reaches the planet's surface. For sustainable energy systems, global solar radiance is regarded as being an essential metric. It was vital to examine the sunlight available in each region before installing any photovoltaic technology. In this study, solar radiation is analyzed with the help of the meteorological dataset of Indian cities (Mumbai, Chennai & Delhi) from meteoblue (meteoblue.com). Comparison is made in two parts; of each town, the data from 2015-2019 is used for training, and data from 2020 is taken for testing. In the second case, data for 2015-2020 for two cities is used for training, and the data for the third city is taken for testing. The independent variables of the dataset include sunshine duration, month, cloud cover, soil temperature, mean monthly temperature, 2m temperature (air temperature at 2 meters above the surface), sea level pressure, wind speed, max temp, min temp. First, independent variables are used for the regression model, and then a stepwise Multi Non-Linear Regression (MNLR) model is applied to find the optimal input variables. Three classifiers were applied to find and compare the performance of the models, namely Artificial Neural Networks (ANN), Support Vector Machine (SVM), and Recurrent Neural Network (RNN), which is a deep neural network mode. The accuracy results of all the SVM, MNLR, ANN, and RNN were comparable, but ANN gave better results with Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE)

Browse

Recent Submissions