Machine learning (ML) has become a core feature for today's real-world applications, making it trending topic the software engineering community. Requirements Engineering (RE) is no stranger to this and its main conferences have included workshops aiming at discussing RE in context of ML. However, current research on intersection between ML mainly focuses using techniques support activities rather than exploring how can improve development ML-based systems. This paper concerns systematic...

[Context] Applying design principles has long been acknowledged as beneficial for understanding and maintainability in traditional software projects. These benefits may similarly hold Machine Learning (ML) projects, which involve iterative experimentation with data, models, algorithms. However, ML components are often developed by data scientists diverse educational backgrounds, potentially resulting code that doesn't adhere to best practices. [Goal] In order better understand this...

[Context] Technical debt (TD) in machine learning (ML) systems, much like its counterpart software engineering (SE), holds the potential to lead future rework, posing risks productivity, quality, and team morale. Despite growing attention TD SE, understanding of ML-specific code-related remains underexplored. [Objective] This paper aims identify discuss relevance issues that ML code throughout workflow. [Method] The study first compiled a list 34 contributing by examining phases workflow,...

This work presents a Genetic Fuzzy Classification System (GFCS) called Programming (GPF-CLASS). model differs from the traditional approach of GFCS, which uses metaheuristic as way to learn "if-then" fuzzy rules. classical needs several changes and constraints on use genetic operators, evaluation selection, depends primarily used. makes this implementation costly explores few its characteristics potentialities. The GPF-CLASS seeks for greater integration with metaheuristic: Multi-Gene...

The relevance of business process modeling and information technology is growing. That happens because, in general, systems automate, assist control those processes. So, the understanding can be crucial for an appropriate requirements definition. However, there are still few methods that exploit contributions Business Process Engineering provide elicitation to produce results more compliant company needs. Thus, this work presents a method its application real case uses processes as way...

[Context] In Brazil, 41% of companies use machine learning (ML) to some extent. However, several challenges have been reported when engineering ML-enabled systems, including unrealistic customer expectations and vagueness in ML problem specifications. Literature suggests that Requirements Engineering (RE) practices tools may help alleviate these issues, yet there is insufficient understanding RE’s practical application its perception among practitioners. [Goal] This study aims investigate...

This work describes the use of a weighted ensemble neural network classifiers for adaptive learning. We train networks by means quantum-inspired evolutionary algorithm (QIEA). The QIEA is also used to determine best weights each classifier belonging when new block data arrives. After running several simulations using two different datasets and performing analysis results, we show that proposed algorithm, named neuro-evolutionary (NEVE), was able learn set quickly respond any drifts on...

In our previous works [1, 2], we proposed NEVE, a model that uses weighted ensemble of neural network classifiers for adaptive learning, trained by means quantum-inspired evolutionary algorithm (QIEA). We showed the neuro-evolutionary were able to learn dataset and quickly respond any drifts on underlying data. Now, are particularly interested analyzing influence an unlimited ensemble, instead limited from NEVE. For that, modified NEVE work with ensembles, call this new NEVE++. To verity how...

Abstract This work presents a new neuro-evolutionary model, called NEVE (Neuroevolutionary Ensemble), based on an ensemble of Multi-Layer Perceptron (MLP) neural networks for learning in nonstationary environments. makes use quantum-inspired evolutionary models to automatically configure the members and combine their output. The identify most appropriate topology each MLP network, select relevant input variables, determine network weights calculate voting weight member. Four different...

Combining forecasts is a common practice in time series analysis. This technique involves weighing each estimate of different models order to minimize the error between resulting output and target. work presents novel methodology, aiming combine using genetic programming, metaheuristic that searches for nonlinear combination selection forecasters simultaneously. To present method, authors made three tests comparing with linear forecasting combination, evaluating both terms RMSE MAPE. The...

Software is strategic for Brazil's development, but the lack of a larger qualified workforce limits country's productive capacity. This paper reports experience deploying large-scale distance learning education program to meet practical needs software industry. We applied design thinking involving engineering and continued experts PUC-Rio an external consultancy methodology that we consider ideal, including objects hitherto not used in Brazil. The key elements conceived include (i)...

This work presents a Genetic Fuzzy Controller (GFC), called Programming Inference System for Control tasks (GPFIS-Control). It is based on Multi-Gene Programming, variant of canonical Programming. The main characteristics and concepts this approach are described, as well its distinctions from other GFCs. Two benchmarks application GPFIS-Control considered: the Cart-Centering Problem Inverted Pendulum. In both cases results demonstrate superiority potentialities in relation to GFCs found literature.

This work describes the use of a weighted ensemble neural network classifiers for adaptive learning. We train networks by means quantum-inspired evolutionary algorithm (QIEA). The QIEA is also used to determine best weights each classifier belonging when new block data arrives. show that neuroevolutionary are able learn set and quickly respond any drifts on underlying data. compare results reached our model with an existing algorithm, Learn++.NSE, in two different nonstationary scenarios.

Context: The number of TV series offered nowadays is very high. Due to its large amount, many are canceled due a lack originality that generates low audience.

Collaborative evaluation is still a less explored subject within the collaboration field. This paper proposes methodology to collaborative based on 3C model, which can be applied both in learning or working groups. also presents two case studies run graduate courses of Computer Science Department Catholic University Rio de Janeiro for students work. The suggest that appropriate this type assessment. In addition, participating experiment rated experience as positive, confirming premise value...

The materialization of the universalization social protection, foreseen in Constitution Brazil chapter Social Security, with tripod Health, Welfare and Assistance, specifically scope Welfare, occurs through granting maintenance benefits to all Brazilians who need this which generates a huge demand for millions requests annual INSS, is operator these services. Receiving analyzing benefit requests, among other processes, timely manner assertiveness, complex challenging, whether due volume or...

Context: Virtual analyzers or inferences are mathematical models widely used in the chemical process industry, as they allow real-time prediction of properties interest from measurements basic available at appropriate frequency. Problem: Therefore, a key component control and optimization structures. Inferences with superior capabilities provide better reducing losses improving profits. Solution: This work seeks to establish whether Machine Learning (ML) algorithms techniques could be...

[Context] Applying design principles has long been acknowledged as beneficial for understanding and maintainability in traditional software projects. These benefits may similarly hold Machine Learning (ML) projects, which involve iterative experimentation with data, models, algorithms. However, ML components are often developed by data scientists diverse educational backgrounds, potentially resulting code that doesn't adhere to best practices. [Goal] In order better understand this...

Context: Given the significant expenses incurred by General Social Security Regime (RGPS) in disability retirement benefit payments, accurately assessing costs associated with new concessions is crucial for maintaining financial and actuarial balance of system. Entry tables are essential instruments this measurement.

Context: The growing challenge in attracting employees to Voluntary Resignation Programs (VRP) lies the need balance company's cost control with goal of increasing participation from target audience. Problem: It is essential ensure that process occurs smoothly, reducing tension during separation and fostering a more cooperative responsible environment. Companies maximize attraction VRP, minimize costs, improve resource allocation. Solution: This article aims construct Machine Learning (ML)...