3830 research outputs found
Sort by
AI-Driven Spend Analysis Application: Integrating Purchase Order Classification Proactive Procurement Forecasting & Spend Visibility
No full textThis research introduces a groundbreaking advancement in the realm of spend analytics within procurement, presenting an automated solution that integrates cutting-edge machine learning models with Microsoft Power BI. Utilizing Convolutional Neural Networks (CNN) for precise text classification of Purchase Orders (PO) and employing RandomForestClassifier, RandomForestRegressor, XGBClassifier, and XGBRegressor for forecasting both spend and the most procured categories, this methodology constitutes a substantial contribution. The implementation of batch file automation streamlines all process components with a single click. The CNN model enhances efficiency and accuracy by automating the classification of purchase order text, significantly reducing manual efforts in procurement. Simultaneously, the RandomForest and XGBoost models contribute to robust forecasting, delivering proactive insights. The study meticulously details the development, training, and seamless integration of these models within the Power BI environment, offering insights into both challenges and successes. Real-world application and rigorous testing validate the practicality of the solution, demonstrating improved accuracy in text-based purchase order classification and resilient forecasting capabilities. Results indicate a CNN accuracy of 70% for transaction categories, highlighting its adaptability. Forecasting models, particularly XGBoost, exhibit superior accuracy with minimal deviation, achieving a variance of 1.5%. This automated approach transforms spend analysis methodologies. The paper concludes by discussing broader implications, and potential advancements, and suggesting future avenues for refining ensemble machine-learning applications
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
Advancing financial strategy: A comprehensive analysis of ESG and value investing for profit maximization
No full textThis research explores the comparative expected returns between portfolios of value stocks and top-rated ESG (Environmental, Social, and Governance). Our analytical approach included a rigorous selection process for value stocks based on financial metrics and for ESG stocks based on their sustainability scores. Our comprehensive analysis included calculating expected returns, risk assessments, and portfolio optimizations, which were facilitated through advanced financial tools and methodologies.
This research underscores the shifting paradigms in investment strategies, where ethical considerations are increasingly becoming as crucial as financial outcomes. It asks the question can investors align themselves with broader environment and societal goals and still achieve profitable returns and open the path for further research that could extend the dept of the analysis.
The findings advocate for a broader acceptance and integration of ESG factors as a lucrative investment and a forerunner in the future of investment decision-making, supporting the notion that sustainable investing can indeed align with, or even enhance, financial goals
Advancing Parkinson's Disease Monitoring: Developing AI-Enhanced Models for Predicting Disease Severity from Speech Data
Get PDFThis thesis develops machine learning models to predict Parkinson's disease (PD) severity scores from speech recordings. 22 vocal features related to frequency, amplitude, timing, jitter, shimmer etc. are extracted from sustained vowels, words, and sentences in the PD data. After preprocessing, exploratory analysis provides insights into feature distributions and correlations with severity scores like the Movement Disorders Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS). Rigorous feature selection identifies the most predictive subsets. Multiple sophisticated algorithms including random forests, support vector machines and neural networks are benchmarked and tuned using cross-validation to predict the MDS-UPDRS scores. A nonlinear support vector regressor with optimal features achieves high accuracy. Thorough model interpretation explains performance, identifies limitations guiding future improvements and characterizes clinical implementation requirements. Overall, the interpretable modeling approach accurately forecasts PD severity, enabling potential telemonitoring applications
Comparing machine learning algorithm for predicting loan application for performance enahncement
No full textThis study evaluates various machine learning algorithms for predicting bank loan status using the CRISP-DM methodology. The research utilizes a dataset from Kaggle, focusing on algorithms such as Decision Trees, Support Vector Machines, AdaBoost, Random Forests, K-Nearest Neighbours, and Logistic Regression. Models were optimized using GridSearchCV, with recall as the primary metric to highlight the importance of accurately detecting loan repayment patterns.
The findings demonstrate that the Support Vector Machine model was the most effective, achieving a recall score of 0.99and an F1 score of 0.96. Ensemble methods, which combine multiple models, notably improved prediction accuracy while maintaining interpretability. This study identifies the most effective algorithms and provides insights into factors influencing loan decisions, offering practical recommendations for banks to reduce bad loan rates and promote sustainable lending practices through advanced machine learning techniques
Spam Classification Using Machine Learning and Deep Learning
No full textThis research explores a comprehensive approach to refining spam classification accuracy by integrating traditional machine learning and advanced deep learning models. Our evaluation encompasses four diverse models—Adaboost, XGBoost, Long Short-Term Memory (LSTM), Feedforward Neural Network (FFN), and an innovative Transformer-CNN hybrid model. Notably, XGBoost emerges as the frontrunner, achieving a remarkable accuracy of 97.84%, closely trailed by Adaboost at 96.72%. The deep learning counterparts, LSTM and FFN, demonstrate competitive accuracies of 96.47% and 97.67%, respectively. Furthermore, the proposed Transformer-CNN hybrid model exhibits a commendable accuracy of 97.07%. This study underscores the pivotal role of amalgamating diverse machine learning paradigms, emphasizing the efficacy of hybrid models in significantly enhancing accuracy and overall performance in spam classification. The findings contribute valuable insights to the domain, showcasing the potential of a unified approach in fortifying email security and advancing the state-of-the-art in spam detection mechanisms
Impact of prior victimization on susceptibility to phishing attacks
Get PDFThis study evaluated various machine learning (ML) algorithms to predict susceptibility to phishing attacks based on prior victimization. The models assessed included Decision Tree Classifier, Support Vector Classifier (SVC), Random Forest, XGBoost, and Logistic Regression. SVC and Logistic Regression achieved the highest accuracy of 0.86 and an F1 score of 0.79, making them top performers. Random Forest also showed strong results with an accuracy of 0.85 and an F1 score of 0.79, while XGBoost had an accuracy of 0.82 and an F1 score of 0.79. The Decision Tree Classifier was the least effective, with an accuracy of 0.75 and an F1 score of 0.76. Feature selection significantly enhanced model performance, and the
quality and size of the training dataset were crucial. This study concludes that SVC and Logistic Regression are the most effective models for predicting phishing susceptibility, offering valuable insights for improving cybersecurity measures
Employee attrition forecasting: Determining the optimal algorithm for predicting employee turnover
No full textEmployee attrition is a severe organizational challenge contributing to decline in both the productivity and profitability. The research target is to identify the universal machine learning algorithm which could predict employee attrition through examination of diverse factors including employee demographics, termination reasons, and job-related specifications. The study has Logistic Regression, Random Forest, and Gradient Boosting, models that were compared in terms of their functionality in predicting an individual leaving the company. Exploratory Data Analysis (EDA) is intended in the early stages of the analysis when the dataset structure is analyzed and the main factors that cause attrition are identified. Development of models, followed by their evaluation using metrics such as accuracy, precision, recall, F1 score, area under the ROC curve, and precision recall curves takes place concurrently.
The results display that the Gradient Boosting some of the other models in terms of accuracy and recall highlighting the robustness of the model in capturing the imbalanced nature of the dataset. The fundamental condition of the reasons for attrition consist of termination causes, voluntary and involuntary types of termination, age, and length of service. The paper demonstrates the value of these factors in predicting the attrition an in addition it supplies insights into fruitful attrition management techniques.
The survey closes with recommendations for organizations to take an advanced machine learning approach and data-driven strategy in order to reduce staff turnover. Future trends encompass the discovery of the additional predictive indicators, the enhancement of the algorithm models and the implementation of those techniques across different fields to ensure the high-performance capability of the model. This research is a part of the broad field of human resources analytics techniques that confirm the applicability of machine learning to address critical workforce problems
Effect on Performance of Employees Working in Rotational Shifts within the Pharmaceutical Industry in India
No full textThe comprehensive study on rotational shift work in the Indian pharmaceutical industry illuminates the impact of irregular work schedules on employee performance, well-being, and job satisfaction, particularly among younger adults. A key finding is the negative impact of rotational shifts on work-life balance and overall health, leading to stress, sleep disturbances, and long-term health issues. These irregular hours not only affect personal lives but also professional performance, increasing the risk of errors in a precision-essential field. The study advocates for flexible scheduling, such as shift swapping and self-scheduling, to give employees more control over their work hours, enhancing job satisfaction and reducing stress. Additionally, it recommends holistic health and wellness programs, including regular health check-ups, stress management workshops, and mental health support. It also underscores the need for an inclusive, supportive work culture with open communication channels. Implementing these measures can result in a more engaged, productive, and satisfied workforce, which is vital for maintaining high productivity and quality standards in the Indian pharmaceutical industry
The Applications of Data Mining Techniques in Detecting Occupational Fraud: A Qualitative Review of Forensic Accounting Practices
No full textThis study examines the application of data mining techniques to forensic accounting in terms of fraud detection. The research methodology involves qualitative methods, which include case studies and in-depth interviews with veteran forensic accountants concerning how data mining tools are used to perform their work. It then compares the results of such applications to those obtained through traditional forms of expert opinion analysis. It becomes clear from the study that traditional accounting methods are powerful but often inadequate for handling big data and unable to do real-time analysis. One of the most important topics highlighted in this research is deep learning and artificial intelligence, which are providing increasingly complex fraud schemes with powerful tools. Also, it examines the ethical and legal aspects of data mining in fraud detection. This research makes a contribution to the academic debate on forensic accounting and data mining, offering practical advice for improving occupational fraud detection capabilities