ML Framework Research Articles

The Future of Healthcare Data Governance AI-Enabled ETL Testing Frameworks on Data Warehousing Testing automation using ML, AI and NLP

The healthcare sector enhances the integration of artificial intelligence, natural language processing and machine learning in data governance for developing automation procedures. Data management offered AI-powered ETL testing with the adoption of an automation framework for resolving data governance issues . This insight promotes operational efficiency with assurance of data quality that empowers data security by ML, AI and NLP frameworks. The framework allows healthcare institutions to arrange unstructured data in effective insights into patient care and treatment procedures. Advanced analytics leads to the development of predictions on patient outcomes in treatment procedures. The automated process of ETL improves efficient information collection by NPL with the creation of predictive models in monitoring data management by ML

An artificial neural network visible mathematical model for predicting slug liquid holdup in low to high viscosity multiphase flow for horizontal to vertical pipes

AbstractSlug liquid holdup (SLH) is a critical requirement for accurate pressure drop prediction during multiphase pipe flows and by extension optimal gas lift design and production optimization in wellbores. Existing empirical correlations provide inaccurate predictions because they were developed with regression analysis and data measured for limited ranges of flow conditions. Existing SLH machine learning models provide accurate predictions but are published without any equations making their use by other researchers difficult. The only existing ML model published with actual equations cannot be considered optimum because it was selected by considering artificial neural network (ANN) structures with only one hidden layer. In this study, an ANN-based model for SLH prediction with actual equations is presented. A total of 2699 data points randomly divided into 70%, 15%, and 15% for training, validation, and testing was used in constructing 71 different network structures with 1, 2, and 3 hidden layers respectively. Sensitivity analysis revealed that the optimum network structure has 3 hidden layers with 20, 5, and 15 neurons in the first, second, and third hidden layers, respectively. The optimum network structure was translated into actual equations with the aid of the weights, biases, and activation functions. Trend analysis revealed that this study’s model reproduced the expected effects of inputs on SLH. Test against measured data revealed that this study’s model is in agreement with measured data with coefficient of determinations of 0.9791, 0.9727, 0.9756, and 0.9776 for training, testing, validation, and entire datasets, respectively. Comparative study revealed that this study’s model outperformed existing models with a relative performance factor of 0.002. The present model is presented with visible mathematical equations making its implementation by any user easy and without the need for any ML framework. Unlike existing ANN-based models developed with one hidden layered ANN structures, the present model was developed by considering ANN structures with one, two, and three hidden layers, respectively.

Using machine learning to combine genetic and environmental data for maize grain yield predictions across multi-environment trials

Key messageIncorporating feature-engineered environmental data into machine learning-based genomic prediction models is an efficient approach to indirectly model genotype-by-environment interactions.Complementing phenotypic traits and molecular markers with high-dimensional data such as climate and soil information is becoming a common practice in breeding programs. This study explored new ways to combine non-genetic information in genomic prediction models using machine learning. Using the multi-environment trial data from the Genomes To Fields initiative, different models to predict maize grain yield were adjusted using various inputs: genetic, environmental, or a combination of both, either in an additive (genetic-and-environmental; G+E) or a multiplicative (genotype-by-environment interaction; GEI) manner. When including environmental data, the mean prediction accuracy of machine learning genomic prediction models increased up to 7% over the well-established Factor Analytic Multiplicative Mixed Model among the three cross-validation scenarios evaluated. Moreover, using the G+E model was more advantageous than the GEI model given the superior, or at least comparable, prediction accuracy, the lower usage of computational memory and time, and the flexibility of accounting for interactions by construction. Our results illustrate the flexibility provided by the ML framework, particularly with feature engineering. We show that the feature engineering stage offers a viable option for envirotyping and generates valuable information for machine learning-based genomic prediction models. Furthermore, we verified that the genotype-by-environment interactions may be considered using tree-based approaches without explicitly including interactions in the model. These findings support the growing interest in merging high-dimensional genotypic and environmental data into predictive modeling.

ICURE: Intensive care unit (ICU) risk evaluation for 30-day mortality. Developing and evaluating a multivariable machine learning prediction model for patients admitted to the general ICU in Sweden.

A prediction model that estimates mortality at admission to the intensive care unit (ICU) is of potential benefit to both patients and society. Logistic regression models like Simplified Acute Physiology Score 3 (SAPS 3) and APACHE are the traditional ICU mortality prediction models. With the emergence of machine learning (machine learning) and artificial intelligence, new possibilities arise to create prediction models that have the potential to sharpen predictive accuracy and reduce the likelihood of misclassification in the prediction of 30-day mortality. We used the Swedish Intensive Care Registry (SIR) to identify and include all patients ≥18 years of age admitted to general ICUs in Sweden from 2008 to 2022 with SAPS 3 score registered. Only data collected within 1 h of ICU admission was used. We had 153 candidate predictors including baseline characteristics, previous medical conditions, blood works, physiological parameters, cause of admission, and initial treatment. We stratified the data randomly on the outcome variable 30-day mortality and created a training set (80% of data) and a test set (20% of data). We evaluated several hundred prediction models using multiple ML frameworks including random forest, gradient boosting, neural networks, and logistic regression models. Model performance was evaluated by comparing the receiver operator characteristic area under the curve (AUC-ROC). The best performing model was fine-tuned by optimizing hyperparameters. The model's calibration was evaluated by a calibration belt. Ultimately, we simplified the best performing model with the top 1-20 predictors. We included 296,344 first-time ICU admissions. We found age, Glasgow Coma Scale, creatinine, systolic blood pressure, and pH being the most important predictors. The AUC-ROC was 0.884 in test data using all predictors, specificity 95.2%, sensitivity 47.0%, negative predictive value of 87.9% and positive predictive value of 70.7%. The final model showed excellent calibration. The ICU risk evaluation for 30-day mortality (ICURE) prediction model performed equally well to the SAPS 3 score with only eight variables and improved further with the addition of more variables. The ICURE prediction model predicts 30-day mortality rate at first-time ICU admission superiorly compared to the established SAPS 3 score.

An ML-based decision support system for reliable diagnosis of ovarian cancer by leveraging explainable AI

Ovarian cancer (OC) is one of the most prevalent types of cancer in women. Early and accurate diagnosis is crucial for the survival of the patients. However, the majority of women are diagnosed in advanced stages due to the lack of effective biomarkers and accurate screening tools. While previous studies sought a common biomarker, our study suggests different biomarkers for the premenopausal and postmenopausal populations. This can provide a new perspective in the search for novel predictors for the effective diagnosis of OC. Genetic algorithm has been utilized to identify the most significant biomarkers. The XGBoost classifier is then trained on the selected features and high ROC-AUC scores of 0.864 and 0.911 have been obtained for the premenopausal and postmenopausal populations, respectively. Lack of explainability is one major limitation of current AI systems. The stochastic nature of the ML algorithms raises concerns about the reliability of the system as it is difficult to interpret the reasons behind the decisions. To increase the trustworthiness and accountability of the diagnostic system as well as to provide transparency and explanations behind the predictions, explainable AI has been incorporated into the ML framework. SHAP is employed to quantify the contributions of the selected biomarkers and determine the most discriminative features. Merging SHAP with the ML models enables clinicians to investigate individual decisions made by the model and gain insights into the factors leading to that prediction. Thus, a hybrid decision support system has been established that can eliminate the bottlenecks caused by the black-box nature of the ML algorithms providing a safe and trustworthy AI tool. The diagnostic accuracy obtained from the proposed system outperforms the existing methods as well as the state-of-the-art ROMA algorithm by a substantial margin which signifies its potential to be an effective tool in the differential diagnosis of OC.

Inference latency prediction for CNNs on heterogeneous mobile devices and ML frameworks

Due to the proliferation of inference tasks on mobile devices, state-of-the-art neural architectures are typically designed using Neural Architecture Search (NAS) to achieve good tradeoffs between machine learning accuracy and inference latency. While measuring inference latency of a huge set of candidate architectures during NAS is not feasible, latency prediction for mobile devices is challenging, because of hardware heterogeneity, optimizations applied by machine learning frameworks, and diversity of neural architectures. Motivated by these challenges, we first quantitatively assess the characteristics of neural architectures (specifically, convolutional neural networks for image classification), ML frameworks, and mobile devices that have significant effects on inference latency. Based on this assessment, we propose an operation-wise framework which addresses these challenges by developing operation-wise latency predictors and achieves high accuracy in end-to-end latency predictions, as shown by our comprehensive evaluations on multiple mobile devices using multicore CPUs and GPUs. To illustrate that our approach does not require expensive data collection, we also show that accurate predictions can be achieved on real-world neural architectures using only small amounts of profiling data.

Beam Prediction for mmWave V2I Communication Using ML-Based Multiclass Classification Algorithms

Beam management is a key functionality in establishing and maintaining reliable communication in cellular and vehicular networks, and it becomes more critical at millimeter-wave (mmWave) frequencies and for high-mobility scenarios. Traditional approaches consume wireless resources and incur high beam training overheads in finding the best beam pairings, thus necessitating alternative approaches such as position-aided, vision-aided, or, more generally, sensing-aided beam prediction approaches. Current systems are also leveraging artificial intelligence/machine learning (ML) to optimize the beam management procedures; however, the majority of the proposed ML frameworks have been applied to synthetic datasets, leading to overestimated performances. In this work, in the context of vehicle-to-infrastructure (V2I) communication and using the real-world DeepSense6G experimental datasets, we investigate the performance of four ML algorithms on beam prediction accuracy for mmWave V2I scenarios. We compare the performance of K-nearest neighbour (KNN), support vector machine (SVM), decision tree (DT), and naïve Bayes (NB) algorithms on position-aided beam prediction accuracy and related metrics such as precision, recall, specificity, and F1-score. The impacts of different beam codebook sizes and dataset split ratios on five different scenarios’ datasets were investigated, independently and collectively. Confusion matrices and area under the receiver operating characteristic curves were also employed to visualize the (mis)classification statistics of the considered ML algorithms. The results show that SVM outperforms the other three algorithms, for the most part, on the scenario-per-scenario cases. However, for the combined scenario with larger data samples, DT outperforms the other three algorithms for both the different codebook sizes and data split ratios. The results also show comparable performance for the different data split ratios considered for the different algorithms. However, with respect to the codebook sizes, the results show that the higher the codebook size, the lower the beam prediction accuracy. With the best accuracy results around 70% for the combined scenario in this study, multi-modal sensing-aided approaches can be explored to increase the beam prediction performance, although at the expense of higher system complexity when compared to the position-aided approach considered in this study.

Data analysis and machine learning aided integrated catalyst activity and process modelling for selective H2 production from biomass gasification

Hydrogen energy derived through biomass gasification is considered as one of the most sorted sustainable sources of renewable energy. This process enhances the H2 production from biomass in the presence of specific catalysts. Among different kinds of models that have been employed for this process, ML models adept at approximating non-linear functions and facilitate outcome prediction without detailed mathematical descriptions. Thus, the current work focuses on understanding structural-composition-operating-target property relationships, and integrated catalyst and process modelling using ML framework for thermo-catalytic biomass gasification to H2 production, and demonstrates outliers handling, data normalization for efficient handling of data-driven modelling with non-linear database. Linear, tree-based, kernel-based, and ANN models were developed with 589 datapoints screened from the 59 relevant papers with 24 inputs and 4 outputs (H2, CO, CO2, and CH4 as vol. %). Performance of these models are evaluated through 5-fold cross-validation and test data with the help of statistical measures. ANN with Basian-regularization learning algorithm using tan-sigmoid activation function in both layers, resulted superior performance in prediction of H2 production (RMSE = 6.85 & R2 = 0.80) and other output gases with high accuracy (i.e., minimum deviation from experimental data) compared to other ML models. Further, using the best ML model, input contribution and PDP analysis were performed to interpret the significance of predominate input parameters affecting on the product composition. Feature contribution analysis reveals that temperature, S/B ratio, catalyst support type, and sulphur content in biomass are significant parameters for enhancing H2 production from catalytic-biomass gasification, and PDP analysis discloses their optimal operating region.

Forecasting Business Status of Organizations by Analyzing Historic Earnings Call Transcripts with the Aid of Text Refinement Framework

Objectives: The research work focuses on providing an effective framework for automated text refinement that aids in financial condition projections for the company based on prior transcripts of earnings calls. The proposed framework captures the ad hoc advancements of the organizations described in the earnings call as sentiments and computes a score based on the captured sentiments. The sentiment score is then used as prime parameter to predict the stock values of the organizations. Methods: The framework is equipped with sentiment analysis or opinion mining to identify and extract the subjective content using text mining and Natural Language Processing (NLP). The extracted sentiments help in yielding a sentiment score to aid in the process of stock projection. The research also illustrates how the sentiment score-based stock prediction enhances in projections of stock compared to existing ML frameworks like LSTM, Random Forest, ARIMA and Regression models. Findings: The proposed work has an accuracy score of 93%, precision 96% and recall 95% which is comparatively better than existing ML frameworks framed on LSTM, Random Forest, ARIMA and Regression models. Novelty: The research framework overcomes the influence of regular features and test data in stock prediction by using the computed sentiment prediction score from the extracted sentiment phase to aid in prediction stock values and determine the financial status of organizations. The existing frameworks project the stock price based on trained model from previous stock price repository, which tend to fail capturing ad hoc changes incurring in the organization such as change of management or any economic disaster which can poses a high impact on stock projections, the proposed research work captures these organizational changes from the earnings call transcripts as sentiments and build a score to yield the stock projection framework. Keywords: Natural Language Processing, Text Refinement, Loughran McDonald Sentiment Classifier, Term Frequency Inverse Document Frequency, Stock Price

Career Bot for Career Prediction of Higher Secondary Students using Decision Tree

Career guidance refers to a process that assists individuals, typically students in making informed decisions about their career paths. Career guidance can be delivered through various channels, including career counsellors, educational institutions, online platforms, and self-help resources. It plays a vital role in helping individuals make informed choices that align with their aspirations, values, and capabilities. Traditional career prediction models often lack transparency and fail to consider the diverse and dynamic factors that influence students' career choices. The existing systems may exhibit biases and limitations that hinder accurate and personalized career guidance. The project aims to tackle these problems by developing an Explainable ML (XML) model that provides transparent, personalized, and adaptable recommendations to higher secondary students. The proposed system incorporates Decision Tree algorithms within an Explainable ML framework to provide clear and comprehensible insights into the factors influencing career predictions. It takes into account a diverse set of input features, including academic performance, skills, interests, and extracurricular activities, to offer personalized career guidance to individual students. The project also addresses potential biases in the model to ensure fair and equitable career recommendations for students from varied backgrounds. By combining the power of Decision Tree algorithms with Explainable ML, the project aims to empower higher secondary students in making well-informed decisions about their future careers. The transparency provided by the Explainable ML model enhances user trust and understanding, fostering a more engaging and personalized career prediction system. The project's outcomes are expected to contribute significantly to the field of career guidance, providing a model that is not only accurate but also accessible and comprehensible for students navigating the critical phase of choosing their career paths.

Detect the Cardiovascular Disease's in Initial Phase using a Range of Feature Selection Techniques of ML

Heart-related conditions remain the foremost global cause of mortality. In 2000, heart disease claimed around 14 million lives worldwide, a number that surged to approximately 620 million by 2023. The aging and expanding population significantly contribute to this rising mortality trend. However, this also underscores the potential for significant impact through early intervention, crucial for reducing fatalities from heart failure, where prevention plays a pivotal role. The aim of the present research is to develop a prospective ML framework that can detect important features and predict cardiac conditions as an early stage using a variety of choice of features strategies. The Features subsets that were chosen were designated as FST1, FST2, and FST3, respectively. Three distinct methods, including correlation-based feature selection, chi-square and mutual information, were used for picking features. Next, the most confident theory & the most appropriate feature selection were identified using six alternative machine learning models: Logistical Regression (LR) (AL1), the support vector Machine (SVM ) (AL2), K-nearest neighbor (K-NN) (AL3), Random forest (RF) model (AL4), Naive Bayes (NB) model (AL5), and Decision Tree (DT) (AL6). Ultimately, we discovered that, with 95.25% accuracy, 95.11% sensitivity, 95.23% specificity, 96.96 area below receiver operating characteristic and 0.27 log loss, the random forest model offered the most excellent results for F3 feature sets. No one has investigated coronary artery disease forecasting in depth; however, our study evaluates multiple statistics (specificity, sensitivity, accuracy, AUROC, and log loss) and uses multiple attribute choices to improve algorithms success for important features. The suggested model has considerable promise for medical use to speculate CVD find in Precursor at a minimal cost and in a shorter amount of time as well as will assist limited experience physician to take right decision based on the results of the used model combined with specific criteria.

Enhancing Privacy in Machine Learning Services with Hybrid Homomorphic Encryption: Guard ML Framework

Enhancing Privacy in Machine Learning Services with Hybrid Homomorphic Encryption: Guard ML Framework

The nature and practices of the use of Machine Learning and Deep Learning Frameworks to assist Software Project Management: A Developing Country Context

The study conducted a comprehensive analysis of literature and interviews related to the use of ML and DL in SPM, identifying challenges and gaps in the field. The study adopts a qualitative research design, gathering data through semi-structured interviews with project managers involved in SPM using ML and DL frameworks. The collected data is analysed using thematic analysis to capture the participants' subjective experiences, perceptions, and opinions. The findings suggest that while the use of AI techniques for SPM has been gaining attention, it is still in its infancy, with limited research and practice in developing countries. The study also emphasizes the unique challenges faced by developing countries in SPM, including limited resources, diverse stakeholders, cultural differences, and rapid changes. It suggests that ML and DL frameworks can offer advantages in enhancing the productivity, innovation, and competitiveness of software projects in these contexts. However, the study acknowledges that there are limitations to these AI techniques for SPM, and further research is needed to address these limitations. The study bridge the gap between theory and practice by offering practical insights for project managers and recommendations for future research in utilizing ML and DL frameworks to assist SPM.

GeneAI 3.0: powerful, novel, generalized hybrid and ensemble deep learning frameworks for miRNA species classification of stationary patterns from nucleotides

Due to the intricate relationship between the small non-coding ribonucleic acid (miRNA) sequences, the classification of miRNA species, namely Human, Gorilla, Rat, and Mouse is challenging. Previous methods are not robust and accurate. In this study, we present AtheroPoint’s GeneAI 3.0, a powerful, novel, and generalized method for extracting features from the fixed patterns of purines and pyrimidines in each miRNA sequence in ensemble paradigms in machine learning (EML) and convolutional neural network (CNN)-based deep learning (EDL) frameworks. GeneAI 3.0 utilized five conventional (Entropy, Dissimilarity, Energy, Homogeneity, and Contrast), and three contemporary (Shannon entropy, Hurst exponent, Fractal dimension) features, to generate a composite feature set from given miRNA sequences which were then passed into our ML and DL classification framework. A set of 11 new classifiers was designed consisting of 5 EML and 6 EDL for binary/multiclass classification. It was benchmarked against 9 solo ML (SML), 6 solo DL (SDL), 12 hybrid DL (HDL) models, resulting in a total of 11 + 27 = 38 models were designed. Four hypotheses were formulated and validated using explainable AI (XAI) as well as reliability/statistical tests. The order of the mean performance using accuracy (ACC)/area-under-the-curve (AUC) of the 24 DL classifiers was: EDL > HDL > SDL. The mean performance of EDL models with CNN layers was superior to that without CNN layers by 0.73%/0.92%. Mean performance of EML models was superior to SML models with improvements of ACC/AUC by 6.24%/6.46%. EDL models performed significantly better than EML models, with a mean increase in ACC/AUC of 7.09%/6.96%. The GeneAI 3.0 tool produced expected XAI feature plots, and the statistical tests showed significant p-values. Ensemble models with composite features are highly effective and generalized models for effectively classifying miRNA sequences.

Abstract 903: AURIGIN: A comprehensive single-cell OMICs atlas of human development and an AI/ML framework to classify and identify the drivers of tumor plasticity and altered cellular state

Abstract Altered cellular state and plasticity have been increasingly identified as drivers of poor prognosis and treatment-resistance in cancer. Identifying therapeutics that correct altered cellular state and plasticity meets a broad need in the field of oncology, particularly in tumors resistant to standard-of-care therapy. Tumor plasticity and altered cellular states encompass many potential configurations, including dedifferentiated and transdifferentiated tumors, so classifying tumors based on their specific cell state profile is critical to identifying the gene drivers that are specific to that altered state. For that reason, we have developed AURIGIN - a platform that combines a comprehensive single-cell OMICs atlas of human development with an AI/ML framework that can classify cell state plasticity and identify the genes and pathways that drive those specific altered cellular states. AURIGIN is more comprehensive than previously disclosed single cell atlases in that it focuses and enriches the diverse stem, progenitor, and developmental states commonly involved in plastic cell states of cancer. The platform standardizes and integrates these states with multiple atlases of differentiated adult tissues to create a unified atlas of human development. In addition, the AI/ML paradigm of AURIGIN is embedded with developmental biology models that enable it to deconvolute state heterogeneity within tumor indications for precise identification of the most relevant gene targets for novel therapies. AURIGIN also unifies classified plasticity across tissue types by identifying altered cellular states that span multiple indications. Furthermore, we present the implementation of AURIGINDRIVE ML models that integrate pathway information, physical interaction datasets, and regulatory annotations to accurately predict the gene targets at the top of hierarchy of control of tumor plasticity. AURIGIN also enables and accelerates target validation by identifying models that map to the relevant classified plastic states. Similarly, AURIGIN defines clinically relevant patient selection and treatment efficacy biomarkers that are specifically defined from the classified plastic cellular state. We demonstrate the success of AURIGIN for therapeutic discovery. In sum, AURIGIN amplifies the value of multiOMIC tumor data in cellular state and plasticity target discovery by mapping the data into a developmental biology-informed ML framework. Citation Format: Joseph DeBartolo, Henry Wilson, Kimberly S. Straley, Maulasri Bhatta, Sambad Sharma, Sara Sinicropi-Yao, James Neef, Betty Chan, Andrew McRiner, Mark Bittinger, Laura Antipov, Katharine E. Yen, Thomas G. Graeber, David S. Millan. AURIGIN: A comprehensive single-cell OMICs atlas of human development and an AI/ML framework to classify and identify the drivers of tumor plasticity and altered cellular state [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 903.

Machine Learning‐Assisted Property Prediction of Solid‐State Electrolyte

AbstractMachine learning (ML) exhibits substantial potential for predicting the properties of solid‐state electrolytes (SSEs). By integrating experimental or/and simulation data within ML frameworks, the discovery and development of advanced SSEs can be accelerated, ultimately facilitating their application in high‐end energy storage systems. This review commences with an introduction to the background of SSEs, including their explicit definition, comprehensive classification, intrinsic physical/chemical properties, underlying mechanisms governing their conductivity, challenges, and future developments. An in‐depth explanation of the ML methodology is also elucidated. Subsequently, the key factors that influence the performance of SSEs are summarized, including thermal expansion, modulus, diffusivity, ionic conductivity, reaction energy, migration barrier, band gap, and activation energy. Finally, it is offered perspectives on the design prerequisites for upcoming generations of SSEs, focusing on real‐time property prediction, multi‐property optimization, multiscale modeling, transfer learning, automation and high‐throughput experimentation, and synergistic optimization of full battery, all of which are crucial for accelerating the progress in SSEs. This review aims to guide the design and optimization of novel SSE materials for the practical realization of efficient and reliable SSEs in energy storage technologies.

Adversarial Defense: A GAN-IF Based Cyber-security Model for Intrusion Detection in Software Piracy

Software-piracy continues to be most critical distress, posing grave threats to digital-assets and financial stability. Traditional Intrusion Detection systems (IDS) often battles hard to identify latest piracy attempts owing to their dependence on pre-established patterns. To effectively address this we attempt to suggest innovative approach leveraging DL based Generative Adversarial Networks (GANs) and ML based Isolation Forest (IF) for detecting software piracies. Our proposed GAN-IF based cyber-security model performs its functions by training a Generator network to mimic the behavior of genuine software applications. Discriminator network discriminates between legitimate and pirated software. Isolation Forests assists in detecting anomalies in diverse conditions, including unseen attacks. Integrated training based on DL and ML framework enables efficient learning and adaptation with respect to piracy challenges, making it highly-successful against prior known threats. There are several DL models which are utilized in IDS operations having limitations in terms of robustness, interpretability. Utilizing GAN in the context of cyber-security to combat software-piracy can have noteworthy merits since GANs can precisely identify forged software as they are skilled at generating fake content resembling actual. Training a GAN on legitimate software, helps to learn and identify disparities in pirated versions. Isolation-Forest can detect anomalies in software distribution networks or user behavior with respect to software usage by recognizing abnormal patterns indicating software piracy, like illegal access or sharing of software licenses. Our proposed model combines GANs and Isolation Forests, excels at accurately detecting subtle indicators of software piracy, a capability that traditional methods may fail to recognize. ML-DL integrated model continuously learns and updates its detection capabilities in response to evolving piracy tactics, making it resilient against zero-day attacks, polymorphic malware. Through adversarial training, ml-model minimizes false alarms and focuses only on genuine threats. In our evaluation, we demonstrate the effectiveness of GAN-IF based cyber-security model in detecting software piracy attempts across various scenarios. Results indicate that our approach outperforms traditional solutions in terms of detection accuracy and adaptability.

ML-driven risk estimation for memory failure in a data center environment with convolutional neural networks, self-supervised data labeling and distribution-based model drift determination

With the trend towards multi-socket server systems, the demand for random access memory (RAM) per server increased. The consequence are more DIMM sockets per server. Since every dual in-line memory module (DIMM), which comprises a series of dynamic random-access memory integrated circuits, has a probability of failure, RAM issues became a dominant failure pattern for servers. The concept introduced in this work contributes to improving the reliability of data centers by avoiding RAM failures and mitigating their impact. For this purpose, an ML-driven framework is provided to estimate the probability of memory failure for each RAM module. The ML framework is based on structural information between correctable (CE) and uncorrectable errors (UE). In a common memory scenario, a corrupted bit within a module can be restored by redundancy using an error correction code (ECC), resulting in a CE. However, if there is more than one corrupted bit within a group of bits covered by the ECC, the information cannot be restored, resulting in a UE.Consequently, the related task requesting the memory content, and the corresponding service may crash. There is evidence that UEs have a CE history and structural relation between the CEs. However, for the case of UEs without a CE history or of a false decision of the ML framework, we extend the total framework by engineering measures to mitigate the impact of a UE by avoiding kernel panic and using backups. The engineering measures use a mapping between physical and logical memory addresses.

Fairness in Machine Learning Meets with Equity in Healthcare

With the growing utilization of machine learning in healthcare, there is increasing potential to enhance healthcare outcomes. However, this also brings the risk of perpetuating biases in data and model design that can harm certain demographic groups based on factors such as age, gender, and race. This study proposes an artificial intelligence framework, grounded in software engineering principles, for identifying and mitigating biases in data and models while ensuring fairness in healthcare settings. A case study is presented to demonstrate how systematic biases in data can lead to amplified biases in model predictions, and machine learning methods are suggested to pre-vent such biases. Future research aims to test and validate the proposed ML framework in real-world clinical settings to evaluate its impact on promoting health equity.

Pushing ML Predictions Into DBMSs.

In the past decade, many approaches have been suggested to execute ML workloads on a DBMS. However, most of them have looked at in-DBMS ML from a training perspective, whereas ML inference has been largely overlooked. We think that this is an important gap to fill for two main reasons: (1) in the near future, every application will be infused with some sort of ML capability; (2) behind every web page, application, and enterprise there is a DBMS, whereby in-DBMS inference is an appealing solution both for efficiency (e.g., less data movement), performance (e.g., cross-optimizations between relational operators and ML) and governance. In this article, we study whether DBMSs are a good fit for prediction serving. We introduce a technique for translating trained ML pipelines containing both featurizers (e.g., one-hot encoding) and models (e.g., linear and tree-based models) into SQL queries, and we compare in-DBMS performance against popular ML frameworks such as Sklearn and ml.net. Our experiments show that, when pushed inside a DBMS, trained ML pipelines can have performance comparable to ML frameworks in several scenarios, while they perform quite poorly on text featurization and over (even simple) neural networks.