1,721,186 research outputs found
Tailoring biomaterial surface properties to modulate host-implant interactions: implication in cardiovascular and bone therapy
Get PDFHost body response to a foreign medical device plays a critical role in defining its fate post implantation. It is thus important to control host-material interactions by designing innovative implant surfaces. In the recent years, biochemical and topographical features have been explored as main target to produce this new type of bioinert or bioresponsive implants. The review discusses specific biofunctional materials and strategies to achieve a precise control over implant surface properties and presents possible solutions to develop next generation of implants, particularly in the fields of bone and cardiovascular therapy
Mental Illness Detection Using Natural Language Processing and Machine Learning
No full textMental illness nowadays has a huge impact on public health that encourages new methods of early identification and intervention. This dissertation will explore the bridge between Natural Language Processing and Machine Learning methods. The dataset is taken from Zenodo that has 104 files and one million rows and reflects different mental health conditions. Altogether total rows were one million and one hundred thousand is taken randomly from them where each file contributes ten percent of the data. Preprocessing techniques were applied to improve the quality of the train set such as stopword removal, lemmatization, punctuation removal, special character and number removal, and data merging. For model training different feature engineering techniques were used such as TF-IDF, Min-Max scaling, Standard scaling and Log scaling. On the other hand, four different classifiers were used to evaluate the effectiveness of predicting mental diseases from text. They are Multinomial Naive Bayes, Logistic Regression, Random Forest and Gradient Boosting. Grid search and Random search were also be used to investigate the difference between the results of Logistic Regression and Multinomial Naive Bayes. To evaluate the performance of each model different techniques were used like accuracy, weighted precision, F1 scores, and confusion matrices. Logistic Regression was the best model which was min-max scaled
Impact of funding and geographical factors on software startups' success
No full textThe purpose of this study was to investigate and adress two important questions regarding the success of software startups. To begin, it investigated the level of impact that different types of funding structures and geographical factors have on the success of the software startup companies. It explored various machine learning models to predict the outcomes of startup ventures, taking into account important features, model performance and cost-effectiveness. As demonstrated in this report, the research provided answers to these questions.
The study identified the primary factors that contribute to over 64 per cent of the success or failure of a software startup company. Location-region accounts for 18%, Time initial funding has received accounts for 16%, timing of Final Funding accounts for 15%, Access to Venture Capital accounts for 11%, and Location - Country accounts for 5%.
Logistic regression emerged as the most suitable model for deployment, achieving an accuracy of 96.58%, precision of 96.51%, recall of 100%, and an F1 score of 98.22%. This is by utilising the CRISP-DM methodology, Python code, and Power BI for data scrutiny and analysis. In addition, this model provides significant cost savings, which amount to 73.81 million dollars. The study does, however, acknowledge that there are challenges associated with limitations in the dataset scope and timeframe. These findings highlight the significance of conducting a comprehensive analysis of a startup, which should include aspects such as financial evaluation, geographical analysis, and predictive modelling
Enhancing Abstractive and Extractive Reviews Text Summarization using NLP and Neural Networks
No full textIn today's digital age, the exponential growth of online reviews has presented businesses with the challenge of efficiently processing huge volumes of textual data. The ability to filter valuable insights from these reviews is crucial for understanding consumer sentiments and facilitating informed decision-making. This thesis investigates the application of a Seq2Seq Long Short-Term Memory (LSTM) model for text summarization, aiming to develop an automated system capable of generating concise and informative summaries from input texts. The model underwent comprehensive training, comprising three epochs, during which it exhibited considerable progress in reducing loss and improving cosine similarity metrics, signifying its ability to learn from the dataset.
Upon evaluation, the model showcased its potential by generating summaries. However, a critical analysis revealed certain limitations in the generated summaries, notably their brevity without substantive depth and occasional lack of coherence. This highlighted the current challenges in achieving nuanced and contextually appropriate summarizations.
A comparative analysis against established models emphasized the need for further advancements. The model's deficiency in comprehensiveness and contextual relevance, particularly when faced with complex sentence structures or nuanced semantic relations, was evident.
Identified limitations encompassed the depth and complexity of summaries, reliance on a limited dataset size, and struggles with processing intricate linguistic features. To address these challenges and enhance model performance, potential avenues include dataset augmentation with diverse text samples, exploration of advanced architectures such as transformer-based models, and meticulous hyperparameter tuning.
In conclusion, while the Seq2Seq LSTM model exhibits promise for text summarization tasks, its current limitations impede the generation of comprehensive and contextually accurate summaries. The thesis underscores the necessity for future endeavors to mitigate these limitations, thereby advancing the model's capability to produce more informative and contextually apt summaries
Unraveling Emotions in Lyrics: A Novel Approach to Enhance Spotify Music Recommendations
Get PDFMusic recommendation systems play a pivotal role in user engagement, satisfaction, and retention on streaming platforms like Spotify. However, traditional methods often fall short of providing diverse and emotionally resonant song suggestions, leading to repetitive playlists and user dissatisfaction. This research therefore delves into the unexplored area of emotion detection within song lyrics to enhance personalized music recommendations on Spotify. The study investigates various machine learning models, including Logistic Regression, Support Vector Machines, Bidirectional LSTM, and DistilBERT, to understand the intricate emotional hints within song lyrics. Through a comparative evaluation of these models, the research identifies Bidirectional LSTM as the most effective, achieving an accuracy of 92%, followed by Random Forest at 82%, and Support Vector Machines and Decision Tree both at 70%. Additionally, the research examines hybrid recommender systems combining case-based k Nearest Neighbours and content-based filtering to offer users nuanced and emotionally connecting song recommendations. The project seeks to optimize music discovery, boost user engagement, foster industry innovation, and ensure a more inclusive representation of artists and genres. Ultimately, the research aspires to introduce a novel perspective to music recommendation systems, one that authentically resonates with user emotions, preferences, and satisfaction
Comparative study of traditional vs. transformer machine learning algorithms for Inflammatory Bowel Disease (IBD) Medical Report Classification
No full textInflammatory Bowel Disease (IBD) is a serious chronic condition that affects millions worldwide, yet it remains under-researched, especially in the application of machine learning. This research aims to draw attention to IBD and support the efforts of My Chron's and Colitis Angel (MyCCAngel), a non-profit organization, by developing a medical report classification model to determine whether users of their platform have IBD. The study compares traditional machine learning models, including Multi-layer Perceptron, Support Vector Machines, and Naïve Bayes, with a transformer-based model, Bidirectional Encoder Representations from Transformers (BERT).
MyCCAngel offers a social platform specifically designed for IBD patients, providing them with tools and assistance to manage their daily challenges. Transformer models, such as BERT, represent a recent advancement in machine learning, applying an evolving set of mathematical techniques (Merritt, 2022). This research seeks to evaluate whether these newer models outperform traditional methods in the classification of medical reports.
In this study, a quantitative research approach is employed, relying on data collected from a substantial sample size. This method allows for the identification of patterns and trends within the data, providing a more scalable and objective analysis. By leveraging a dataset, the study aims to draw conclusions that are applicable beyond the immediate sample, enhancing the reliability and applicability of the findings.
In this study, a quantitative research approach is employed, involving the manual collection of 110 medical reports, both IBD and non-IBD, due to the unavailability of an existing dataset. This method allows for the identification of patterns and trends within the data, providing a more scalable and objective analysis. By using dataset, the study aims to draw conclusions that are applicable.
The performance of each model was evaluated based on metrics such as accuracy, precision, recall, and F1-score. The findings indicate that traditional machine learning algorithms, particularly Naïve Bayes, outperform the transformer-based model BERT, achieving an accuracy of 91% compared to BERT's 68%.
This study demonstrates that transformer models are not always superior and that traditional simple models like Naïve Bayes can offer better performance in specific tasks, such as IBD medical report classification. Furthermore, this research is the first to focus on the classification of IBD medical reports, providing valuable insights for future binary classification tasks in the medical field
Deciphering Deception - Detecting Fake Review using NLP by analysis of stylistic, sentiment-based, and semantic features
Get PDFThis study delves into the critical issue of identifying deceptive online reviews, a challenge increasingly prevalent in the digital marketplace. The study leverages a combination of Natural Language Processing (NLP) and Machine Learning (ML) techniques to differentiate between genuine and fraudulent reviews. The methodology encompasses stylistic analysis to assess language structure, sentiment analysis to evaluate emotional tone, and semantic analysis employing Word2Vec and Latent Dirichlet Allocation (LDA) to uncover latent topics. These components form the foundation for feature engineering for model training and evaluation.
A diverse range of machine learning models, including Random Forest, Logistic Regression, Gaussian and Multinomial Naive Bayes, Simple Neural Network, Gradient Boosted Trees, Convolutional Neural Network (CNN), and Recurrent Neural Network (RNN) with Long Short- Term Memory (LSTM), were comprehensively evaluated. The comparative analysis provides valuable insights into the performance characteristics of each model.
Notably, Logistic Regression and Simple Neural Network emerge as top contenders, presenting strong accuracy, precision, recall, and F1 score. This comparative study serves as a benchmark for future research in the domain, offering a clear understanding of the strengths and weaknesses of various machine learning approaches in addressing the deceptive online review problem, using the combination of stylistic, sentiment-based, and semantic analysis. This research not only advances the understanding of deceptive review detection but also offers a foundation for future explorations in the field of NLP and ML, aimed at enhancing the reliability and transparency of online consumer feedback
Speech Emotion Recognition Using Deep Learning
No full textThis study delves into the realm of emotion recognition in speech, employing advanced deep learning techniques to analyse and categorize emotions from audio data. The primary dataset used is Crema, a comprehensive collection of vocal expressions representing various emotions. The research involves preprocessing the audio data and extracting meaningful features, particularly Mel-Frequency Cepstral Coefficients (MFCCs) alongside x-vectors, which are crucial in understanding the tonal aspects of the speech. The processed data is then fed into two different neural network models: a Recurrent Neural Network (RNN) with SimpleRNN layers and a Long Short-Term Memory (LSTM) network. These models are trained and validated on the dataset to classify emotions into categories such as neutral, happy, sad, angry, fear, and disgust. The performance of these models is evaluated based on metrics like accuracy and F1 score. Results indicate a significant potential of deep learning in effectively recognizing and categorizing emotions in speech, though challenges in accuracy and model optimization persist. Keywords: Emotion Recognition, Speech Processing, Deep Learning, Neural Networks, MFCC, RNN, LSTM, Audio Data Analysis
Microencapsulated genetically engineered Lactobacillus plantarum 80 (pCBH1) for bile acid deconjugation and its implication in lowering cholesterol
No full textA novel approach whereby one can use genetically engineered cells for the purpose of deconjugating bile acids and lowering cholesterol is presented in this thesis. The concept of oral administration of artificial cells has been used for this purpose. In concurrence to these requirements, several in-vitro methods are designed and discussed in this report. For these studies, alginate-polylysine-alginate microcapsules were used. Process parameters for Lactobacillus plantarum 80 (pCBH1) cell microencapsulation have been presented. Results show the possibility of using these artificial cells for various applications. For in-vitro experiments, immobilized bacteria and artificial cells containing the genetically engineered organism were challenged with physiologically relevant levels of bile acids. Results show that immobilized and microencapsulated genetically engineered bacterial cells are capable of lowering physiological levels of bile acids in-vitro. Further, this report summarizes the physiological interrelationship between bile acids and cholesterol and predicts oral doses of microencapsulated cells required for lowering cholesterol
- …
