Data Preparation Phase Research Articles

Better together: Automated app review analysis with deep multi-task learning

Context:User reviews of mobile apps provide an important communication channel between developers and users. Existing approaches to automated app review analysis mainly focus on one task (e.g., bug classification task, information extraction task, etc.) at a time, and are often constrained by the manually defined patterns and the ignorance of the correlations among the tasks. Recently, multi-task learning (MTL) has been successfully applied in many scenarios, with the potential to address the limitations associated with app review mining tasks. Objective:In this paper, we propose MABLE, a deep MTL-based and semantic-aware approach, to improve app review analysis by exploiting task correlations. Methods:MABLE jointly identifies the types of involved bugs reported in the review and extracts the fine-grained features where bugs might occur. It consists of three main phases: (1) data preparation phase, which prepares data to allow data sharing beyond single task learning; (2) model construction phase, which employs a BERT model as the shared representation layer to capture the semantic meanings of reviews, and task-specific layers to model two tasks in parallel; (3) model training phase, which enables eavesdropping by shared loss function between the two related tasks. Results:Evaluation results on six apps show that MABLE outperforms ten commonly-used and state-of-the-art baselines, with the precision of 79.76% and the recall of 79.24% for classifying bugs, and the precision of 79.83% and the recall of 80.33% for extracting problematic app features. The MTL mechanism improves the F-measure of two tasks by 3.80% and 4.63%, respectively. Conclusion:The proposed approach provides a novel and effective way to jointly learn two related review analysis tasks, and sheds light on exploring other review mining tasks.

PollenNet: A novel architecture for high precision pollen grain classification through deep learning and explainable AI

Pollen grains play a critical role in environmental, agricultural, and allergy research despite their tiny dimensions. The accurate classification of pollen grains remains a significant challenge, mainly attributable to their intricate structures and the extensive diversity of species. Traditional methods often lack accuracy and effectiveness, prompting the need for advanced solutions. This study introduces a novel deep learning framework, PollenNet, designed to tackle the intricate challenge of pollen grain image classification. The efficiency of PollenNet is thoroughly evaluated through stratified 5-fold cross-validation, comparing it with cutting-edge methods to demonstrate its superior performance. A comprehensive data preparation phase is conducted, including removing duplicates and low-quality images, applying Non-local Means Denoising for noise reduction, and Gamma correction to adjust image brightness. Furthermore, Explainable AI (XAI) is utilized to enhance the interpretability of the model, while Receiver Operating Characteristic (ROC) curve analysis serves as a quantitative method for evaluating the model's capabilities. PollenNet demonstrates superior performance when compared to existing models, with an accuracy of 98.45 %, precision of 98.20 %, specificity of 98.40 %, recall of 98.30 %, and f1-score of 98.25 %. The model also maintains low Mean Squared Error (0.03) and Mean Absolute Error (0.02) rates. The ROC curve analysis, the low False Positive Rate (0.016), and the False Negative Rate (0.017) highlight the reliability and dependability of the model. This study significantly improves the efficacy of classifying pollen grains, indicating an important advancement in the application of deep learning for ecological research.

A deep learning framework for predicting slab transverse crack using multivariate LSTM-FCN in continuous casting

Accurate and timely predictions of transverse cracks in slabs are crucial for ensuring efficient and high-quality production in continuous casting. However, the accumulation of substantial amounts of complex, multi-dimensional time series data during the casting process considerably affects the modeling performance. In order to increase the prediction accuracy, this paper introduces a novel prediction framework for slab transverse cracks (PF-STC), which comprises three stages: data preparation, data processing, and model training. First, a synchronized scheme is developed for real-time data tracking to achieve one-to-one matching between the process data in the data preparation phase. Second, the data preprocessing stage tackles the multidimensionality of time series data through normalization and mean aggregation downsampling. Finally, in the model training phase, a multivariate long short-term memory-fully convolutional network (MLSTM-FCN) algorithm is utilized to improve the predictive performance. Experimental results on a real dataset from a steel manufacturing company demonstrate that the PF-STC framework outperforms recent state-of-the-art methods, reducing the number of parameters by 49.97% and training time by 33.83%. The PF-STC achieves an AUC score of 0.9435, with accuracy, precision, recall, and F1 scores of 0.9163, 0.8657, 0.8286, and 0.8467, respectively. The proposed PF-STC is poised to provide support for online transverse crack monitoring, with the application demands of monitoring production processes for the steel industry.

Total electricity generation dynamics analysis and renewable energy impacts in South Africa

AbstractThis research explores the dynamics of total electricity generation (TEG) in South Africa through an analysis of data from the International Energy Agency database from 1990 to 2020. A comprehensive examination of various energy sources, including coal, oil, biofuels, nuclear, hydro, solar photovoltaic (PV), solar thermal, and wind, is conducted to ascertain their respective contributions to TEG. Employing the R software environment, the study employs a methodical analytical framework encompassing meticulous data preparation, statistical analysis, and model formulation. The data preparation phase involves intricate processes such as structuring, cleansing, and visualization aimed at eliminating stochastic variables and outliers. Missing data are addressed through the application of the Piecewise Cubic Hermite Interpolating Polynomial method. Subsequent statistical analyses are informed by tests for normality and homogeneity of variance, revealing deviations from normality and disparate variances across energy source groups. Consequently, non‐parametric methodologies such as the Kruskal–Wallis test are adopted. Findings underscore the significant role of nuclear energy in TEG despite facing challenges. Model development entails the construction of multiple linear regression models with varying predictor sizes, with Model m06 emerging as the optimal choice, incorporating key predictors such as coal, nuclear, and solar PV. Rigorous diagnostic assessments confirm the robustness of Model m06 and its suitability for TEG prediction. Comparative analysis against actual data validates its superior performance, characterized by minimal errors and high predictive accuracy. The efficacy of Model m06 in capturing TEG dynamics underscores its utility for informing energy planning initiatives. Recommendations derived from the study advocate for prioritizing renewable energy integration, infrastructure investment, research endeavors, monitoring mechanisms, and public awareness campaigns to advance sustainable energy development goals in South Africa.

A framework for modelling customer invoice payment predictions

By offering clients attractive credit terms on sales, a company may increase its turnover, but granting credit also incurs the cost of money tied up in accounts receivable (AR), increased administration and a heightened probability of incurring bad debt. The management of credit sales, although eminently important to any business, is often performed manually, which may be time-consuming, expensive and inaccurate. Such an administrative workload becomes increasingly cumbersome as the number of credit sales increases. As a result, a new approach towards proactively identifying invoices from AR accounts that are likely to be paid late, or not at all, has recently been proposed in the literature, with the aim of employing intervention strategies more effectively. Several computational techniques from the credit scoring literature and particularly techniques from the realms of survival analysis or machine learning have been embedded in the aforementioned approach. This body of work is, however, lacking due to the limited guidance provided during the data preparation phase of the model development process and because survival analytic and machine learning techniques have not yet been ensembled. In this paper, we propose a generic framework for modelling invoice payment predictions with the aim of facilitating the process of preparing transaction data for analysis, generating relevant features from past customer behaviours, and selecting and ensembling suitable models for predicting the time to payment associated with invoices. We also introduce a new sequential ensembling approach, called the Survival Boost algorithm. The rationale behind this method is that features generated by a survival analytic model can enhance the efficacy of a machine learning classification algorithm.

Transformer-inspired stacked-GAN for hyperspectral target detection

ABSTRACT Hyperspectral images (HSI) can provide abundant spectral feature information and have been successfully used in the field of object detection. The use of Generative Adversarial Network (GAN) to reconstruct the background spectral to achieve differential detection has shown excellent performance in hyperspectral object detection (HTD) tasks. However, due to factors such as noise interference and the complexity of the background, such algorithms also face many challenges such as lack of background sample mixing, poor reconstruction effect, and generation of out-of-class samples. A Transformer-inspired stacked GAN network (STGAN) is proposed to solve these problems. First, in the data preparation phase, we adopted a method of extracting pure background samples at a deep level. Then, in the encoder module, the background spectral features are mined using a Transformer-Encoder with Position Coding and Multihead Self-Attention Mechanism (MHSA). Finally, the same GAN is stacked based on the above network structure, and the latent feature mapping of the first GAN is used as input to reduce the out-of-class samples and achieve a more detailed reconstruction of the background. The comparison experiments of our model with other algorithms on datasets in different scenarios prove its advanced performance.

A many objective based feature selection model for software defect prediction

SummaryGiven the escalating magnitude and intricacy of software systems, software measurement data often contains irrelevant and redundant features, resulting in significant resource and storage requirements for software defect prediction (SDP). Feature selection (FS) has a vital impact on the initial data preparation phase of SDP. Nonetheless, existing FS methods suffer from issues such as insignificant dimensionality reduction, low accuracy in classifying chosen optimal feature sets, and neglect of complex interactions and dependencies between defect data and features as well as between features and classes. To tackle the aforementioned problems, this paper proposes a many‐objective SDPFS (MOSDPFS) model and the binary many‐objective PSO algorithm with adaptive enhanced selection strategy (BMaOPSO‐AR2) is proposed within this paper. MOSDPFS selects F1 score, the number of features within subsets, and correlation and redundancy measures based on mutual information (MI) as optimization objectives. BMaOPSO‐AR2 constructs a binary version of MaOPSO using transfer functions specifically for binary classification. Adaptive update formulas and the introduction of the R2 indicator are employed to augment the variety and convergence of algorithm. Additionally, performance of MOSDPFS and BMaOPSO‐AR2 are tested on the NASA‐MDP and PROMISE datasets. Numerical results prove that a proposed model and algorithm effectively reduces feature count while enhancing predictive accuracy and minimizing model complexity.

Optimized long short-term memory with rough set for sustainable forecasting renewable energy generation

Research and development in the field of renewable energy is receiving more attention as a result of the growing demand for clean, sustainable energy. This paper proposes a model for forecasting renewable energy generation. The proposed model consists of three main phases: data preparation, feature selection-based rough set and nutcracker optimization algorithm (NOA), and data classification and cross-validation. First, the missing values are tackled using the mean method. Then, data normalization and data shuffling are applied in the data preparation phase. In the second phase, a new feature selection algorithm is proposed based on rough set theory and NOA, namely RSNOA. The proposed RSNOA is based on adopting the rough set method as the fitness function during the searching mechanism to find the optimal feature subset. Finally, a custom long-short-term memory architecture with the k-fold cross-validation method is utilized in the last phase. The experimental results revealed that the proposed model is very competitive. It is achieved with 4.2113 root mean square error, 0.96 R², 2.835 mean absolute error, and 4.6349 mean absolute percentage error. The findings also show that the proposed model has great promise as a useful tool for accurately forecasting renewable energy generation across various sources.

Advancing Opthalmic Diagnostics: U-Net for Retinal Blood Vessel Segmentation

This research project focuses on the development and evaluation of an advanced algorithm for retinal vessel segmentation, a critical component in the automated analysis of retinal images for diagnosing ocular diseases. Leveraging state-of-the-art image processing techniques and deep learning models, we propose a novel segmentation algorithm that significantly enhances the accuracy and efficiency of identifying retinal blood vessels from fundus photographs. Our methodology encompasses a comprehensive data preparation phase, including image normalization and augmentation, to improve the model's robustness and generalizability. We implemented a convolutional neural network (CNN)-based architecture optimized for the intricate patterns and variations inherent in retinal images. The performance of our algorithm was rigorously evaluated against established benchmarks, demonstrating superior precision, recall, and a higher Dice coefficient compared to existing methods. These findings indicate the potential of our approach to contribute substantially to the early detection and monitoring of ocular conditions such as diabetic retinopathy and glaucoma. The research underscores the importance of advanced computational techniques in enhancing the diagnostic capabilities of retinal image analysis and sets the stage for future innovations in medical imaging.

Feature selection for sentiment analysis using hybrid multiobjective evolutionary algorithm

As the volume of data continues to grow, the significance of text classification is on the rise. This vast amount of data majorly exists in the form of texts. Effective data preparation is essential to extract sentiment data from this vast amount of text, as irrelevant and redundant information can impede valuable insights. Feature selection is an important step in the data preparation phase as it eliminates irrelevant and insignificant features from the huge features set. There exist a large body of work related to feature selection for image processing but limited research is done for text data. While some studies recognize the significance of feature selection in text classification, but there is still need for more efficient sentiment analysis models that optimize feature selection and reduce computational. This manuscript aims to bridge these gaps by introducing a hybrid multi-objective evolutionary algorithm as a feature selection mechanism, combining the power of multiple objectives and evolutionary processes. The approach combines two feature selection techniques within a binary classification model: a filter method, Information Gain (IG), and an evolutionary wrapper method, Binary Multi-Objective Grey Wolf Optimizer (BMOGWO). Experimental evaluations are conducted across six diverse datasets. It achieves a reduction of over 90 percent in feature size while improving accuracy by nearly nine percent. These results showcase the model’s efficiency in terms of computational time and its efficacy in terms of higher classification accuracy which improves sentiment analysis performance. This improvement can be beneficial for various applications, including recommendation systems, reviews analysis, and public opinion observation. However, it’s crucial to acknowledge certain limitations of this study. These encompass the need for broader classifier evaluation, and scalability considerations with larger datasets. These identified limitations serve as directions for future research and the enhancement of the proposed approach.

Crop monitoring using remote sensing land use and land change data: Comparative analysis of deep learning methods using pre-trained CNN models

In the context of the rapidly evolving climate dynamics of the early twenty-first century, the interplay between climate change and biospheric integrity is becoming increasingly critical. The pervasive impact of climate change on ecosystems is manifested not only through alterations in average environmental conditions and their variability but also through ancillary shifts such as escalated oceanic acidification and heightened atmospheric CO2 levels. These climatic transformations are further compounded by concurrent ecological stressors, including habitat degradation, defaunation, and fragmentation. Against this backdrop, this study delves into the efficacy of advanced deep learning methodologies for the classification of land cover from satellite imagery, with a particular emphasis on agricultural crop monitoring. The study leverages state-of-the-art pre-trained Convolutional Neural Network (CNN) architectures, namely VGG16, MobileNetV2, DenseNet121, and ResNet50, selected for their architectural sophistication and proven competence in image recognition domains. The research framework encompasses a comprehensive data preparation phase incorporating augmentation techniques, a thorough exploratory data analysis to pinpoint and address class imbalances through the computation of class weights, and the strategic fine-tuning of CNN architectures with tailored classification layers to suit the specificities of land cover classification challenges. The models' performance was rigorously evaluated against benchmarks of accuracy and loss, both during the training phase and on validation datasets, with preventative strategies against overfitting, such as early stopping and adaptive learning rate modifications, being integral to the methodology. The findings illuminate the considerable potential of leveraging pre-trained deep learning models for remote sensing in agriculture, demonstrating that advanced CNN architectures, particularly DenseNet121 and ResNet50, are notably effective in enhancing crop type classification accuracy from satellite imagery. This study contributes valuable insights to the field of precision agriculture, advocating for the integration of sophisticated image recognition technologies to bolster crop monitoring efficacy, thereby enabling more nuanced agricultural decision-making and resource allocation.

SWAT + input data preparation in a scripted workflow: SWATprepR

Input data collection, quality assurance and preparation are central but time_consuming steps in environmental modeling. Errors due to manual processing of model input data can result in an incorrect representation of an environmental system and may consequently lead to implausible model simulations. Correct input data preparation and thorough quality check at an early stage of the model setup procedure are essential to build confidence in model simulation results. Typically, in environmental model applications, many steps in the input data preparation phase have to be repeated with the inflow of new, additional or corrected data. In this study, we selected the widely used SWAT + ecohydrological model as an illustrative example to investigate challenges related to input data preparation. To assist in these tasks, we developed an R package named SWATprepR, which provides functions for typical and repeating SWAT + model input data preparation tasks. The package supports the preparation of weather input files, atmospheric deposition, soil parameters, crop rotations, and observed (control or calibration) data, to name a few, presently with focus on European applications. The SWATprepR functions are integrated in R script workflows and can help SWAT + modelers to avoid repetitive tasks, secure reproducibility and transparently document the data processing steps. Application of the package is illustrated with a test case of a SWAT + model for a small catchment in central Poland.

Detection of malicious URLs using machine learning

AbstractThe detection of fraudulent URLs that lead to malicious websites using addresses similar to those of legitimate websites is a key form of defense against phishing attacks. Currently, in the case of Internet of Things devices is especially relevant, because they usually have access to the Internet, although in many cases they are vulnerable to these phishing attacks. This paper offers an overview of the most relevant techniques for the accurate detection of fraudulent URLs, from the most widely used machine learning and deep learning algorithms, to the application, as a proof of concept, of classification models based on quantum machine learning. Starting from an essential data preparation phase, special attention is paid to the initial comparison of several traditional machine learning models, evaluating them with different datasets and obtaining interesting results that achieve true positive rates greater than 90%. After that first approach, the study moves on to the application of quantum machine learning, analysing the specificities of this recent field and assessing the possibilities it offers for the detection of malicious URLs. Given the limited available literature specifically on the detection of malicious URLs and other cybersecurity issues through quantum machine learning, the research presented here represents a relevant novelty on the combination of both concepts in the form of quantum machine learning algorithms for cybersecurity. Indeed, after the analysis of several algorithms, encouraging results have been obtained that open the door to further research on the application of quantum computing in the field of cybersecurity.

A Green Fingerprint of Antarctica: Drones, Hyperspectral Imaging, and Machine Learning for Moss and Lichen Classification

Mapping Antarctic Specially Protected Areas (ASPAs) remains a critical yet challenging task, especially in extreme environments like Antarctica. Traditional methods are often cumbersome, expensive, and risky, with limited satellite data further hindering accuracy. This study addresses these challenges by developing a workflow that enables precise mapping and monitoring of vegetation in ASPAs. The processing pipeline of this workflow integrates small unmanned aerial vehicles (UAVs)—or drones—to collect hyperspectral and multispectral imagery (HSI and MSI), global navigation satellite system (GNSS) enhanced with real-time kinematics (RTK) to collect ground control points (GCPs), and supervised machine learning classifiers. This workflow was validated in the field by acquiring ground and aerial data at ASPA 135, Windmill Islands, East Antarctica. The data preparation phase involves a data fusion technique to integrate HSI and MSI data, achieving the collection of georeferenced HSI scans with a resolution of up to 0.3 cm/pixel. From these high-resolution HSI scans, a series of novel spectral indices were proposed to enhance the classification accuracy of the model. Model training was achieved using extreme gradient boosting (XGBoost), with four different combinations tested to identify the best fit for the data. The research results indicate the successful detection and mapping of moss and lichens, with an average accuracy of 95%. Optimised XGBoost models, particularly Model 3 and Model 4, demonstrate the applicability of the custom spectral indices to achieve high accuracy with reduced computing power requirements. The integration of these technologies results in significantly more accurate mapping compared to conventional methods. This workflow serves as a foundational step towards more extensive remote sensing applications in Antarctic and ASPA vegetation mapping, as well as in monitoring the impact of climate change on the Antarctic ecosystem.

Towards Benchmarking for Evaluating Machine Learning Methods in Detecting Outliers in Process Datasets

Within the integration and development of data-driven process models, the underlying process is digitally mapped in a model through sensory data acquisition and subsequent modelling. In this process, challenges of different types and degrees of severity arise in each modelling step, according to the Cross-Industry Standard Process for Data Mining (CRISP-DM). Particularly in the context of data acquisition and integration into the process model, it can be assumed with a sufficiently high degree of probability that the acquired data contain anomalies of various kinds. The outliers must be detected in the data preparation and processing phase and dealt with accordingly. If this is sufficiently implemented, it will positively impact the subsequent modelling in terms of accuracy and precision. Therefore, this paper shows how outliers can be identified using the unsupervised machine learning methods autoencoder, Density-Based Spatial Clustering of Applications with Noise (DBSCAN), Isolation Forest (iForest), and One-Class Support Vector Machine (OCSVM). Following implementing these methods, we compared them by applying the Numenta Anomaly Benchmark (NAB) and sufficiently presented the individual strengths and disadvantages. Evaluating the correctness, distinctiveness and robustness criteria described in the paper showed that the One-Class Support Vector Machine was outstanding among the methods considered. This is because the OCSVM achieved acceptable anomaly detections on the available process datasets with comparatively little effort.

Solution gas-oil ratio estimation using histogram gradient boosting regression, machine learning, and mathematical models: a comparative analysis

ABSTRACT Quantifying gas solubility in oil under reservoir conditions is essential in reservoir engineering. Several laboratory techniques have been suggested to quantify it; nevertheless, these procedures are characterized by their demanding time requirements and high costs. Estimating the solution gas-oil ratio relies significantly on applying mathematical and artificial intelligence models. The research commenced by collecting 757 datasets from published articles. Subsequently, the data was partitioned into two distinct categories: training and test data. Following this data preparation phase, artificial intelligence models, including Histogram Gradient Boosting Regression (HGBoost), Adaptive Boosting (AdaBoost), Support Vector Regression (SVR), and the Adaptive Neuro-Fuzzy Inference System (ANFIS), were employed to model the solution gas-oil ratio. Additionally, the Memetic algorithm (MA) was employed to recompute the coefficients of the Petrosky and Farshad model, initially developed with 81 data points, using an expanded dataset of 605 training instances. Subsequently, model accuracy and performance were assessed through statistical and visual analyses. The results indicate that the HGBoost model performs best, achieving an overall R2 of 0.9886 and an overall RMSE of 41.29 scf/STB. Notably, among the mathematical models, the modified Petrosky and Farshad model stands out with an overall R2 of 0.9078 and an overall RMSE of 114.49 scf/STB. The Taylor diagram and violin plot both indicate that the HGBoost model and the modified Petrosky and Farshad model outperform other machine learning models and mathematical models, respectively. These results demonstrate the accuracy of the models described in this study.

The multifunctional LBC-Corpora: different aims depending on the user

Abstract Corpora are nowadays the primary source of many dictionaries and the core element of various platforms and information systems. Lexicographers have therefore a variety of new possibilities which were unthinkable in the past with different types of corpora available for use in the data collection phase of the lexicographic process (Flinz 2021). Not only lexicographers, but also other types of users (academics, translators, teachers, students etc.) can profit from them, especially when corpora are public and can be accessed using corpus linguistic tools (Ballestracci/Buffagni/Flinz 2020; Flinz/Farina 2020). The LBC-corpora are monolingual specialized comparable corpora, already online (http://corpora.lessicobeniculturali.net/) and, as monitor corpora (Lemnitzer/Zinsmeister 2015: 140), they will be augmented over time. They can be analysed using the open source tool NoSketchEngine (Billero 2020). The LBC-Corpora are also the lexicographic primary source of the LBC multilingual dictionary, which is in preparation: the provisional entry lists of different languages (Spanish, German, and French) are now ready (Billero/Farina/Nicolás Martínez 2020) and together with a selection of KWICS, which have been carefully selected following a quantitative-qualitative procedure (for German see Buffagni/Flinz/Ballestracci in prep.), will soon be online (Flinz et al. in prep.). The purpose of this paper is to reflect on the LBC-corpora from a double perspective: from the user of the LBC-platform and from the lexicographic team. In the first case following an overview of the principal characteristics of the LBC-Platform, the focus will be on the accessible corpora showing the tools which can be used (§ 2). In the second case the LBC-corpora will be examined in their function as a data basis for the LBC-Dictionary (§ 3). The attention will be on the data preparation phase: after discussing the procedure for the realization of the LBC-provisional lemma candidate lists, the focus will be on the adopted procedure for finding equivalence relations and for the individuation of other types of relations between the entries (synonymy, belonging to the same semantic field etc.). In § 4 the focus will be on the LBC-provisional lemma candidate lists and their related KWICs. Conclusions and an outlook to the future can be found in the last section (§ 5).

A Data-centric AI Framework for Automating Exploratory Data Analysis and Data Quality Tasks

Democratisation of machine learning (ML) has been an important theme in the research community for the last several years with notable progress made by the model-building community with automated machine learning models. However, data play a central role in building ML models and there is a need to focus on data-centric AI innovations. In this article, we first map the steps taken by data scientists for the data preparation phase and identify open areas and pain points via user interviews. We then propose a framework and four novel algorithms for exploratory data analysis and data quality for AI steps addressing the pain points from user interviews. We also validate our algorithms with open source datasets and show the effectiveness of our proposed methods. Next, we build a tool that automatically generates python code encompassing the above algorithms and study the usefulness of these algorithms via two user studies with data scientists. We observe from the first study results that the participants who used the tool were able to gain 2× productivity and 6% model improvement over the control group. The second study is performed in a more realistic environment to understand how the tool would be used in real-world scenarios. The results from this study are coherent with the first study and show an average of 30–50% of time savings that can be attributed to the tool.

Uncovering Bias: Exploring Machine Learning Techniques for Detecting and Mitigating Bias in Data – A Literature Review

The presence of Bias in models developed using machine learning algorithms has emerged as a critical issue. This literature review explores the topic of uncovering the existence of bias in data and the application of techniques for detecting and mitigating Bias. The review provides a comprehensive analysis of the existing literature, focusing on pre-processing techniques, post-pre-processing techniques, and fairness constraints employed to uncover and address the existence of Bias in machine learning models. The effectiveness, limitations, and trade-offs of these techniques are examined, highlighting their impact on advocating fairness and equity in decision-making processes.
 The methodology consists of two key steps: data preparation and bias analysis, followed by machine learning model development and evaluation. In the data preparation phase, the dataset is analyzed for biases and pre-processed using techniques like reweighting or relabeling to reduce bias. In the model development phase, suitable algorithms are selected, and fairness metrics are defined and optimized during the training process. The models are then evaluated using performance and fairness measures and the best-performing model is chosen. The methodology ensures a systematic exploration of machine learning techniques to detect and mitigate bias, leading to more equitable decision-making.
 The review begins by examining the techniques of pre-processing, which involve cleaning the data, selecting the features, feature engineering, and sampling. These techniques play an important role in preparing the data to reduce bias and promote fairness in machine learning models. The analysis highlights various studies that have explored the effectiveness of these techniques in uncovering and mitigating bias in data, contributing to the development of more equitable and unbiased machine learning models. Next, the review delves into post-pre-processing techniques that focus on detecting and mitigating bias after the initial data preparation steps. These techniques include bias detection methods that assess the disparate impact or disparate treatment in model predictions, as well as bias mitigation techniques that modify model outputs to achieve fairness across different groups. The evaluation of these techniques, their performance metrics, and potential trade-offs between fairness and accuracy are discussed, providing insights into the challenges and advancements in bias mitigation. Lastly, the review examines fairness constraints, which involve the imposition of rules or guidelines on machine learning algorithms to ensure fairness in predictions or decision-making processes. The analysis explores different fairness constraints, such as demographic parity, equalized odds, and predictive parity, and their effectiveness in reducing bias and advocating fairness in machine learning models. Overall, this literature review provides a comprehensive understanding of the techniques employed to uncover and mitigate the existence of bias in machine learning models. By examining pre-processing techniques, post-pre-processing techniques, and fairness constraints, the review contributes to the development of more fair and unbiased machine learning models, fostering equity and ethical decision-making in various domains. By examining relevant studies, this review provides insights into the effectiveness and limitations of various pre-processing techniques for bias detection and mitigation via Pre-processing, Adversarial learning, Fairness Constraints, and Post-processing techniques.

BiLSTM for Predicting Post-Construction Subsoil Settlement under Embankment: Advancing Sustainable Infrastructure

The load and settlement histories of stage-constructed embankments provide critical insights into long-term surface behavior under embankment loading. However, these data often remain underutilized in predicting post-construction settlement in the absence of geotechnical subsoil characterization. To address this limitation, the current study integrates bidirectional long short-term memory (BiLSTM) into a three-phase framework: data preparation, model construction, and performance evaluation. In the data preparation phase, the feature vector comprises basal pressure, pressure increments, time intervals, and prior settlement values to facilitate a rolling forecast. To manage unevenly spaced data, an Akima spline standardizes the desired time intervals. The model’s efficacy is validated using observational data from two distinct construction case studies, each featuring diverse soil conditions. BiLSTM proves effective in identifying key attributes from load and settlement data during the staged construction process. Compared to traditional curve-fitting methods, the BiLSTM model exhibits superior performance, robustness, and adaptability to varying soil conditions. Additionally, the model demonstrates low sensitivity to the range of post-construction data, allowing for a data collection period reduction—from six months to three—without compromising prediction accuracy (relative error = 0.92%). These advantages not only optimize resource allocation but also contribute to broader sustainability objectives.