Evaluating the Performance of Topic Modeling Techniques with Human Validation to Support Qualitative Analysis

Sensitivity evaluation of machine learning-based calibrated transportation mode choice models: A case study of Alexandria City, Egypt

Systematic Review of Machine Learning in Diagnosis of Myeloproliferative Neoplasia

Data is one of the most important limiting factors of deep machine learning (ML) model in drilling applications. Though a big size of historical data can be available, high-quality cleaned and labeled data is usually limited. In this case study, we show that with limited labeled data, physics-based data augmentation combined with unsupervised learning significantly improves both stability and accuracy in bit wear ML model. It provides a pathway to overcome labeled data shortage and field data quality limitations. Labeled bit wear data is usually limited because only the final bit dull state can be labeled using dull photos for the entire drilling distance. To overcome this difficulty, an encoder-decoder unsupervised ML framework based on Bi-directional LSTM architecture is first applied to the data of entire drilling distance to extract and maximize data learnings. Then, a physics-based wear estimator is implemented in the learned latent space to augment labeled wear dataset, guided by true wear labels and offset-well data. Weighting factor in loss function is applied to optimize augmented and true datasets in final supervised learning step. The proposed real-time bit wear model is built on a combination of physics-based data augmentation and unsupervised data learning method. The model is applied on multiple field bit runs. Results show that by implementing the unsupervised data learnings only, the prediction accuracy is improved by 30% compared to the baseline ML model. By combining physics-based augmentation, the accuracy is further improved by 10%. More importantly, adding physics-based data augmentation significantly reduces prediction variance and unphysical wear outputs, therefore improving prediction stability by more than 30%. It should be noted that the prediction stability of AI/ML model is crucial in real-time application and decision making. The results show physics-based data augmentation not only increases the size of label dataset and prevents model overfitting, but also applies physics-based guidance to the ML model effectively. It is learned weighting factor plays a crucial role in balancing loss contributions from true wear labels and physics-based labels. While low weighting factor of physics-based labels diminishes the augmented data, high weighting factor disrespects the true wear labels leading to a high prediction bias. Overall, the bit wear model study shows physic-based data augmentation combined with unsupervised data learning can effectively improve model accuracy, stability, and overcome labeled data shortage difficulty. The proposed paper shows a case study for the bit wear ML model using a combination of physics-based data augmentation and unsupervised data learning. While labeled data is one of the major challenges of ML models in many drilling applications, this study provides a pathway to improve both accuracy and stability of deep ML models with limited labeled data.

For the various climatic zones of India, machine learning (ML) models are created in the current work to forecast monthly-average diffuse solar radiation (DSR). The long-term solar radiation data are taken from Indian Meteorological Department (IMD), Pune, provided for 21 cities that span all of India’s climatic zones. The diffusion coefficient and diffuse fraction are the two groups of ML models with dual input parameters (sunshine ratio and clearness index) that are built and compared (each category has seven models). To create ML models, two well-known ML techniques, random forest (RF) and k-nearest neighbours (KNN), are used. The proposed ML models are compared with well-known models that are found in the literature. The ML models are ranked according to their overall and within predictive power using the Global Performance Indicator (GPI). It is discovered that KNN models generally outperform RF models. The results reveal that in diffusion coefficient models perform well than diffuse fraction models. Moreover, functional form 2 is the best followed by form 6. The ML models created here can be effectively used to accurately forecast DSR in various climates.

Systematic literature review of machine learning based software development effort estimation models

Machine learning (ML) techniques play a significant role in detecting abnormal link operations in wireless networks. However, during the design and development process of a ML-based model, numerous factors including input data transformation, scaling methods, choice of ML techniques and their associated parametrization, influence the final model performance. In this paper, we discuss the design and development process of wireless link anomaly detection models followed by the model ranking process for selecting the most suitable models given the required application criteria. Under this premise, we consider four different data representations, four anomaly types, six scaling methods and six distinct ML techniques along with their associated parametrization trials leading to over twenty thousand anomaly detection models for wireless links. We show that the most suitable supervised ML models are able to detect anomalies with an F1 score of at least 88% and up to perfect score, whereas unsupervised ML models can attain an F1 score of up to 96% depending on the anomaly type, and on average, amongst 20,352 available models, the supervised SVM model can be readily adopted to efficiently detect any type of anomaly with an F1 score of at least 94%, particularly when trained on encoded data representations using the combination of "mean and deviation scaling" and "non-linear RBF kernel".

Coastal aquifers are vital for sustaining ecosystems and providing freshwater for agriculture and domestic use. However, rising groundwater salinity in these regions demands innovative approaches for accurate modeling and prediction. In recent decades, machine learning (ML) and deep learning (DL) techniques have been extensively used across various water resource management fields, often yielding promising outcomes. This review examines the growing usage of ML and DL techniques in modeling groundwater salinity in coastal aquifers. The analysis is based on 104 peer-reviewed journal articles including the review articles indexed in PubMed, ScienceDirect, Google Scholar, Scopus, and SpringerLink using keywords like "saltwater intrusion", "machine learning", “deep learning”, and "groundwater salinity". The review discusses recent advancements, challenges, and future directions for improving ML model accuracy and applicability. Studies reveal an increasing reliance on ML techniques to predict groundwater salinity and manage seawater intrusion across various coastal geological settings. Researchers have employed a range of ML models, from traditional methods like Artificial Neural Networks and Adaptive Neuro Fuzzy Inference Systems to advanced DL-based models like Convolutional Neural Networks and Long Short-Term Memory Networks, often using ensemble methods to enhance accuracy. Key findings include the top-tier output of ensemble and optimized models, including Random Forest and Grasshopper Optimization Algorithm-XGBoost. These studies highlight the importance of model selection, input variable ranking, and computational efficiency, demonstrating ML's global relevance in environmental management. The review highlights the prospects of ML, DL, and ensemble models to redefine groundwater management in coastal aquifers, offering more effective and sustainable solutions to the challenges posed by groundwater salinity and saltwater intrusion. Future advancements will focus on further integrating DL, explainable artificial intelligence, and ensemble approaches, along with other innovative techniques, to explore sparsely studied variables, model new or unique study areas, and apply ML methods for managing groundwater salinity.

Accurate leaf area index estimation in sorghum using high-resolution UAV data and machine learning models

Smart cities are centres of economic opulence and hope for standardized living. Understanding the shades of urban resilience and livability in smart city models is of paramount importance. This study presents a novel two-stage data-driven framework combining a multivariate metric-distance analysis with machine learning (ML) techniques for resilience and livability assessment of smart cities. A longitudinal dataset for 35 top-ranked European smart cities from 2015 till 2020 applied as the case study under the proposed framework. Initially, a metric distance-based weighting approach is used to weight the indicators and quantify the scores across each aspect under city resilience and urban livability. The key aspects under city resilience include social, economic, infrastructure and built environment and, institutional resilience, while under urban livability, the aspects include accessibility, community well-being, and economic vibrancy. Fuzzy c-means clustering as an unsupervised machine learning technique is used to sort smart cities based on the degree of performance. In addition, an intelligent approach is presented for the prediction of the degree of livability, resilience, and aggregate performance of smart cities based on various supervised ML techniques. Classification models such as Naïve Bayes, k-nearest neighbors (kNN), support vector machine (SVM), Classification and Regression Tree (CART) and, ensemble models including Random Forest (RF) and Gradient Boosting machine (GBM) were used. Three coefficients (accuracy, Cohen's Kappa (κ) and average area under the precision-recall curve (AUC-PR)) along with confusion matrix were used to appraise the performance of the classifier ML models. The results revealed GBM as the best classification and predictive model for the resilience, livability, and aggregate performance assessment. The study also revealed Copenhagen, Geneva, Stockholm, Munich, Helsinki, Vienna, London, Oslo, Zurich, and Amsterdam as the smart cities that co-create resilience and livability in their development model with superior performance.

Summary A meticulous interpretation of steady-state or unsteady-state relative permeability (Kr) experimental data is required to determine a complete set of Kr curves. In this work, different machine learning (ML) models were developed to assist in a faster estimation of these curves from steady-state drainage coreflooding experimental runs. These ML algorithms include gradient boosting (GB), random forest (RF), extreme gradient boosting (XGB), and deep neural network (DNN) with a main focus on and comparison of the two latter algorithms (XGB and DNN). Based on existing mathematical models, a leading-edge framework was developed where a large database of Kr and capillary pressure (Pc) curves were generated. This database was used to perform thousands of coreflood simulation runs representing oil-water drainage steady-state experiments. The results obtained from these simulation runs, mainly pressure drop along with other conventional core analysis data, were used to estimate analytical Kr curves based on Darcy’s law. These analytically estimated Kr curves along with the previously generated Pc curves were fed as features into the ML model. The entire data set was split into 80% for training and 20% for testing. The k-fold cross-validation technique was applied to increase the model’s accuracy by splitting 80% of the training data into 10 folds. In this manner, for each of the 10 experiments, nine folds were used for training and the remaining fold was used for model validation. Once the model was trained and validated, it was subjected to blind testing on the remaining 20% of the data set. The ML model learns to capture fluid flow behavior inside the core from the training data set. In terms of applicability of these ML models, two sets of experimental data were needed as input; the first was the analytically estimated Kr curves from the steady-state drainage coreflooding experiments, while the other was the Pc curves estimated from centrifuge or mercury injection capillary pressure (MICP) measurements. The trained/tested model was then able to estimate Kr curves based on the experimental results fed as input. Furthermore, to test the performance of the ML model when only one set of experimental data is available to an end user, a recurrent neural network (RNN) algorithm was trained/tested to predict Kr curves in the absence of Pc curves as an input. The performance of the three developed models (XGB, DNN, and RNN) was assessed using the values of the coefficient of determination (R2) along with the loss calculated during training/validation of the model. The respective crossplots along with comparisons of ground truth vs. artificial intelligence (AI)-predicted curves indicated that the model is capable of making accurate predictions with an error percentage between 0.2% and 0.6% on history-matching experimental data for all three tested ML techniques. This implies that the AI-based model exhibits better efficiency and reliability in determining Kr curves when compared to conventional methods. The developed ML models by no means replace the need to conduct drainage coreflooding or centrifuge experiments but act as an alternative to existing commercial platforms that are used to interpret experimental data to predict Kr curves. The two main advantages of the developed ML models are their capability of predicting Kr curves within a matter of a few minutes as well as with limited intervention from the end user. The results also include a comparison between classical ML approaches, shallow neural networks, and DNNs in terms of accuracy in predicting the final Kr curves. The research presented here is an extension of the state-of-the-art framework proposed by Mathew et al. (2021). However, the two main aspects of the current study are the application of deep learning for the prediction of Kr curves and the application of feature engineering. The latter not only reduces the training/testing time for the ML models but also enables the end user to obtain the final predictions with the least set of experimental data. The various models discussed in this research work currently focus on the prediction of Kr curves for drainage steady-state experiments; however, the work can be extended to capture the imbibition cycle as well.

A suboptimal management or system malfunction can often lead to abnormal energy consumptions in buildings, which result in a significant waste of energy. For this reason, the adoption of advanced monitoring systems, based on Machine Learning (ML) and visualization techniques, is crucial to avoid possible deviations from the baseline energy consumption. However, the historical data on which analyses are based generally do not report the occurrence of anomalies. Therefore, the application of supervised ML techniques is limited and unsupervised approaches are favored. Moreover, domain experts find most Machine Learning (ML) techniques hard to interpret, and thus find it difficult to contextualize anomalies. To overcome these issues, this work proposes a machine learning-based Anomaly Detection Framework (ADF) that involves the use of two complementary semi-supervised ML applications to obtain a highly interpretable and accurate detection of anomalies. Both techniques use Symbolic Aggregate approXimation (SAX) encoding to extract the most relevant information from load profiles. The aim of the first approach is to maximize the interpretability of the definition and distinction between anomalous and normal behavior. This is achieved using a Classification And Regression Tree (CART), albeit at the expense of a coarser output granularity. The second approach exploits an Multi-Layer Perceptron (MLP) algorithm to obtain a higher and more accurate output resolution, although it leads to a less interpretable definition of any anomalous behavior. The ADF has been applied to a real case study using electricity consumption data provided by a large telecommunications service provider. The results show that combining both ML models enhances the accuracy and interpretability of the detected anomalies.

Security of networks is becoming progressively more necessary as the commercial internet keeps on developing. Intrusion detection techniques using network data have emerged as a significant area of investigation among the various network protection solutions. Many techniques for detecting network anomalies were suggested up to this point. A substantial amount of labeled data is needed for modeling in Machine Learning (ML) that constitutes the basis of the majority of more effective performance detection strategies. Nevertheless, manually filtering and labeling large-scale traffic data is not feasible in a real network. For real detection, network managers can only employ unsupervised ML techniques and the results of such strategies are far less accurate than those of supervised ML strategies. To increase network security’s efficiency of processing, adaptability, and specificity of detection, this method is useful for instantaneous, network-based safety applications because it maximizes intrusion detection accuracy across several attack types by utilizing quantum computer technology. This paper suggests an innovative multiclassification fusion strategy for optimizing network intrusion detection performance. Numerous fusion approaches are used in this research to combine the identification outcomes of effective intrusion detection approaches. A variety of publicly available KDD Cup 99 datasets were utilized for comparative investigations. According to the experiments, the suggested fusion approach [Decision Tree, K-Nearest Neighbor (KNN), improved Siam Intrusion Detection System (I-SiamIDS), Long Short-Term Memory (LSTM)] outperforms individual intrusion detection techniques in multiple measurements. It also exhibits enhanced resilience and stability, making it suitable for use in a variety of network-based safety recognition settings. The study uses the four primary classes of attacks such as Remote-to-Local (R2L), User to Root (U2R), Denial of Service (DoS), and Probing (Probe), which were compared with the performance metrics such as Precision, F-score, Accuracy, and Recall.

Heart disease (HD) remains one of the leading causes of mortality worldwide, emphasizing the urgent need for accurate and timely diagnostic methods. Recent advances in machine learning (ML) have shown promising potential in improving clinical decision-making through predictive modeling. This study aimed to enhance HD prediction by applying selected ML algorithms with hyperparameter optimization to improve their accuracy and reliability. Four classifiers: Decision Tree (DT), Support Vector Machine (SVM), Random Forest (RF), and K-Nearest Neighbors (KNN) were implemented, and missing data were handled using the SimpleImputer technique to ensure robust preprocessing. Model performance was evaluated using standard metrics, including accuracy, precision, recall, F1-score, and the Receiver Operating Characteristic (ROC) curve. The experimental results demonstrated that all classifiers improved after hyperparameter tuning, with KNN achieving the highest accuracy of 96.0%, precision of 0.99, recall of 0.93, F1-score of 0.96, and ROC of 0.98. Comparative analysis revealed that the adopted approach outperformed existing methods reported in related literature, thereby confirming the effectiveness of hyperparameter optimization combined with preprocessing techniques in HD prediction. The study contributes to knowledge by highlighting the significance of optimized ML models in healthcare applications and recommends further validation with larger, diverse datasets for clinical integration.

Summary Downhole vibrations while drilling surface holes could pose significant operational risks, such as premature failure of drillstring components. Using expensive downhole sensors is the most accurate and reliable technique to monitor downhole vibrations in real time. The high cost, however, hinders the operator to use such sensors in each well. This research aims to build different machine learning (ML) models, namely, K-nearest neighbors (K-NN), logistic regression (LR), and random forests (RF), to determine the severity of downhole vibrations (i.e., lateral and torsional while drilling surface holes). Actual field data of 10 wells, where the operator used downhole sensors to monitor downhole vibrations, were used to build predictive models. The input variables included drilling surface parameters and daily operational drilling reports, while the output variables were the severity of the downhole vibrations. One-hot encoding was applied to all categorical variables to convert them to a suitable format for ML modeling. During the model’s development, a new approach that considers the previous actual and predicted vibrations was applied to enable the ML models to learn from the past to classify the severity of vibrations in the future. The developed models were optimized by tuning specific hyperparameters for each technique to reach the optimum models’ performance. Well 11 was used to validate the developed models. The analysis showed that the data set of Well 11 is not balanced in terms of the severity of vibrations. Therefore, the performance of the models was evaluated using macroaveraging and microaveraging methods to depict the actual performance of the models in all classes of the severity of vibrations. The results showed that the LR model outperformed the other models in predicting the severity of downhole lateral and torsional vibrations in Well 11. This indicates that there is a stronger relationship between the input variables and the output variables in our data set in the case of the LR model compared with other models. The approach described in this paper shows potential for continuous monitoring of downhole vibrations in the subsequent offset wells without the need for installing downhole sensors, hence reducing the overall cost of the well.

Aims This research paper aims to check the effectiveness of a variety of machine learning models in classifying esophageal cancer through MRI scans. The current study encompasses Convolutional Neural Network (CNN), K-Nearest Neighbor (KNN), Recurrent Neural Network (RNN), and Visual Geometry Group 16 (VGG16), among others which are elaborated in this paper. This paper aims to identify the most accurate model to facilitate increased, improved diagnostic accuracy to revolutionize early detection methods for this dreadful disease. The ultimate goal is, therefore, to improve the clinical practice performance and its results with advanced machine learning techniques in medical diagnosis. Background Esophageal cancer poses a critical problem for medical oncologists since its pathology is quite complex, and the death rate is exceptionally high. Proper early detection is essential for effective treatment and improved survival. The results are positive, but the conventional diagnostic methods are not sensitive and have low specificity. Recent progress in machine learning methods brings a new possibility to high sensitivity and specificity in the diagnosis. This paper explores the potentiality of different machine-learning models in classifying esophageal cancer through MRI scans to complement the constraints of the traditional diagnostics approach. Objective This study is aimed at verifying whether CNN, KNN, RNN, and VGG16, amongst other advanced machine learning models, are effective in correctly classifying esophageal cancer from MRI scans. This review aims at establishing the diagnostic accuracy of all these models, with the best among all. It plays a role in developing early detection mechanisms that increase patient outcome confidence in the clinical setting. Methods This study applies the approach of comparative analysis by using four unique machine learning models to classify esophageal cancer from MRI scans. This was made possible through the intensive training and validation of the model using a standardized set of MRI data. The model’s effectiveness was assessed using performance evaluation metrics, which included accuracy, precision, recall, and F1 score. Results In classifying esophageal cancers from MRI scans, the current study found VGG16 to be an adequate model, with a high accuracy of 96.66%. CNN took the second position, with an accuracy of 94.5%, showing efficient results for spatial pattern recognition. The model of KNN and RNN also showed commendable performance, with accuracies of 91.44% and 88.97%, respectively, portraying their strengths in proximity-based learning and handling sequential data. These findings underline the potential to add significant value to the processes of esophageal cancer diagnosis using machine learning models. Conclusion The study concluded that machine learning techniques, mainly VGG16 and CNN, had a high potential for escalated diagnostic precision in classifying esophageal cancer from MRI imaging. VGG16 showed great accuracy, while CNN displayed advanced spatial detection, followed by KNN and RNN. Thus, the results set new opportunities for introducing advanced computational models to the clinics, which might transform strategies for early detection to improve patient-centered outcomes in oncology.