Revista de Informática Teórica e Aplicada

Power Allocation in Orthogonal Frequency Division Multiplexing-based Wireless Networks Using Reinforcement Learning

Hudson Henrique de Souza Lopes — 2024-09-04

Energy efficiency is one of the most essential requirements for wireless sensor networks and the internet of things, since batteries are usually the main source of power for these networks. Appropriate techniques based on artificial intelligence can reduce power consumption and provide the required quality of service parameters for the network. In this paper, we approach the challenging problem of allocation of signal transmission power based on the orthogonal frequency division multiplexing technique. We propose to use reinforcement learning-based algorithms to find the optimal policy for allocating power to wireless network devices using a reward function. More specifically, we propose to use the double deep Q-network (DDQN) agent due to its higher learning capacity compared to the Q-learning, deep Q-network (DQN), and the distributed algorithm (DA), a classic algorithm in the literature for power allocation. Simulation results show that the DDQN agent presents promising solutions for power allocation in wireless networks.

A Random Generation of Haskell Programs Applied to Optimization Testing in Compilers

Guilherme Graeff — 2024-09-04

Compilers and interpreters are essential for developing any system or application, so validating their functionality and properties is crucial. These tools are susceptible to failures, like any software artifact, which can introduce errors into programs developed using them. For example, a compiler or interpreter error that alters a program’s behavior can compromise a critical system, and the system impact can be costly. The literature reports several problems found in compilers and interpreters of different languages. When bugs are detected in early releases, they can be reported to the development team, who can fix them before the end user notices the problem. This work describes the procedure used to generate random Haskell programs, which serve as input for property-based tests. More specifically, this work aims to test the compilation and behavior properties of a program, comparing the compilation and execution results of programs generated with different levels of optimization of the Glasgow Haskell Compiler (GHC). From developing a tool that automates the tests, 10,000 random programs were generated, compiled, and executed, of which 57 presented compilation errors with different optimization levels. This result demonstrates that the used approach is promising for error detection and can be improved in further studies.

Performance Evaluation and Operational Availability of a Textile Manufacturing 4.0 Using Stochastic Petri Nets

Edson Moura Silva — 2024-09-04

The Brazilian clothing industry is the 7th largest globally, generating an annual revenue of USD 10 billion. However, Brazil is only 54th in industrial innovation investments. The constraint on industrial innovation investments represents a challenge for the implementation of Industry 4.0, especially in local garment manufacturing hubs. It is crucial to assess the impact of Industry 4.0 in these hubs in order to highlight its benefits. The study proposes the development of an approach based on Stochastic Petri Nets to model production processes in the textile industry, applied to a company in the garment manufacturing hub of the Agreste region of Pernambuco. The results demonstrate that this approach has the potential to assist garment administrators in optimizing production processes and making more informed decisions. Additionally, the approach allows administrators to maintain resources balanced, efficient, and available in accordance with the production goals of the modern market.

Framework for Security Risk Assessment (FSRA) and Fuzzy Risk Inference System (FRIS) based on Standard ISO/IEC 27002:2022

Diógenes Antonio Marques José — 2024-09-04

Information security is a critical aspect for organizations, and a problem in this regard consists of
knowing the level of vulnerability in which organizations. Therefore, this article aims to present a Cybersecurity
Framework (FSRA) based on the ISO/IEC 27002:2022 standard. In addition, a Fuzzy Risk Inference System
(FRIS) is presented, which uses FSRA controls - Security Practices (SP), Software (S), and People (P) as
input, a total of 93 sub-items. A FRIS output is the Security Risk (SR) in which the organization is due to
non-compliance or partial compliance with FSRA controls. Thus, if each control is fully met, its entry value in the
FRIS will be 100%. Otherwise, it will be proportional. Therefore, when the FRIS returns the SR, whose output
values can be Low (13% to 39.99%), Medium (40% to 59.99%), or High (60% to 100%), the security analyst can
measure the risk security, in which the organization is located, according to ISO/IEC 27002:2022 and from that
know where to act and how to act.

Comparative Study of Predictive Models Based on Moving Averages Applied in Binary Options in the Financial Market

Luccas R. Hoff — 2024-09-04

The technological market has been through significant advances in the last few years, and its application in different markets and business lines was driven by the COVID-19 pandemic. In this context, given the global economic crisis that was generated, some ways of earning money online were popularized, including the Binary Options market, which promises high gains but represents a volatile and risky financial niche. The aim is to evaluate the efficiency and assertiveness of these models applied in operations, in addition to analyzing the risks and possible returns. The aim is to provide a more in-depth understanding of the effectiveness of these models, considering the volatile and unpredictable characteristics of the Binary Options financial market. To achieve the proposed objective, different statistical models based on moving averages will be used to compare individual results and between possible combinations of the best classified. This study is expected to contribute to the understanding of the effectiveness of predictive models based on moving averages applied in Binary Options. The results obtained may offer valuable insights for investors and traders interested in this kind of negotiation, helping them to make informed decisions. Furthermore, this work is expected to stimulate the development of more accurate and efficient investment strategies, increasing the probability of success in financial markets. Finally, this study is expected to advance knowledge in the field of financial market forecasting, highlighting the importance of moving averages as a powerful predictive tool in Binary Options.

UnbIAs - Google Translate Gender Bias Mitigation and Addition of Genderless Translation

Vanessa Schenkel — 2024-09-04

Machine Learning, increasingly present in everyday life, is subject to bias. These biases not only reflect social inequalities but also reinforce them. The present study seeks to mitigate gender bias in Google Translate, the most used translation system in the world. For this, we created a translation model with high gender accuracy and performed a linguistic analysis with the spaCy tool and entity identification with roBERTa. The Constrained Beam Search technique is used to maintain the sentence structure of the business model, but with the replacement for the correct genre indicated by the created model. The final sentence is the result of an alignment done with the SimAlign tool. In addition, the present study also produces an algorithm so that sentences without gender indication in English present translations with inflection for feminine, masculine, and neutral gender. Our approach yields a BLEU score of 48.39. In relation to Google Translate, the model increased gender accuracy from 68.75 to 70.09, enhanced in 15.7% the score that measures the difference in accuracy between male and female entities, and improved stereotyped translations in 43%.

Supervised Learning Algorithms Evaluation on Sweet Potato Data for Production Indices Considering Human Consumption

Mardem Arantes de Castro — 2024-09-04

The sweet potato holds significant importance as a root crop cultivated globally, serving purposes in both human and animal nutrition, ethanol fuel production, and ornamental cultivation. Several genetic experiments have been undertaken to identify superior root varieties. Within this study, a refined statistical methodology has been devised to mitigate environmental factors and promote equitable comparisons among diverse varieties. Through the integration of supplementary environmental data into the experimental framework, a deeper understanding of each variety can be attained. This research signifies an inaugural endeavor towards scrutinizing post-harvest sweet potato data, aiming to unveil novel insights.

Ontology for Healthcare AI Privacy in Brazil

Tiago Andres Vaz — 2024-09-04

This article details the creation of a novel domain ontology at the intersection of epidemiology,

medicine, statistics, and computer science. It outlines a systematic approach to handling structured data

anonymously in preparation for its use in Artificial Intelligence (AI) applications in healthcare. The development

followed 7 steps, including defining scope, selecting knowledge, reviewing important terms, constructing classes

that describe designs used in epidemiological studies, machine learning paradigms, types of data and attributes,

risks that anonymized data may be exposed to, privacy attacks, techniques to mitigate re-identification, privacy

models, and metrics for measuring the effects of anonymization. It concludes with a practical implementation of

this ontology in hospital settings to develop and validate AI systems.

Analysis of Forensic Tools for Recovery of Formatted Data: a case study with Microsoft Word files

Rubens Karman Paula da Silva — 2024-09-04

Deleting or formatting files to hide a crime can be considered a frustrating action, given the ease of using forensic software that implements data carving techniques. This research aims to evaluate the accuracy of forensic data carving software when subjected to recovering formatted Microsoft Word files. The software chosen is widely used in the field and has been featured in scientific papers: Foremost, Scalpel, Recurva, PhotoRec, Autopsy and Magic Rescue. The metrics analyzed were: software execution time, number and size of files recovered, number of false positives and true positives generated in three test scenarios. Validation took place by comparing the resulting files with the originals using a hash algorithm. To structure the test scenarios, a dataset was built with 16,000 copies of files of various lengths. In each scenario, the number of files and the requirements that the software was subjected to varied, with only doc or docx files being recovered. Of the software analyzed, Recuva, Autopsy and PhotoRec had the highest percentages of true positives (>90%) in all the scenarios evaluated. As for false positives, Recuva performed better than the others, with approximately 1%.

Rough Sets and Multilayer Perceptron in Tandem for Processing the Aleatory Uncertainty in COVID-19 Cases

Kalyl Henings — 2024-09-04

It is common sense that challenging clinical cases can occur in the practice of medicine. These clinical cases can lead to undesirable situations of diagnostic uncertainty, making it important to identify these difficult cases and lead them to a discussion by experts for appropriate characterization. In remote/isolated regions, it is crucial to have a computational system that can support the health personnel in identifying these challenging cases. Events such as the COVID-19 pandemic have demonstrated the need for this type of system to assist in screening medical exams. Although the chest X-ray examination is a valuable diagnostic tool for COVID-19 cases, some conditions can be so challenging that doctors can be faced with uncertainty in diagnosis. This article proposes a system that combines machine learning via Multilayer Perceptron neural network in conjunction with the detection of uncertainty via Rough Sets in modeling a system that incorporates uncertainty to produce a classification of cases as positive for the disease, negative for the disease, or diagnostically uncertain. This system would serve as support for the rapid and efficient triage of cases, particularly those classified as “uncertain,” for a medical committee of specialists in a video conference, for example. Experiments were carried out and the trained model achieved 87.61% overall accuracy and a hit rate consistent across all classes.

Systematic Literature Review on the Application of Controllers Based on Cellular Automata in Robotic Tasks

Naya Letícia Batista Souza — 2024-09-04

In recent years, robotics has garnered significant attention from researchers, driven by the desire to develop robots capable of undertaking challenging or laborious tasks traditionally performed by humans. Cellular automata (CA) has emerged as an intriguing solution in this realm, albeit with relatively limited exploration in robotics. Nonetheless, its evolution over time is evident through increasing research interest, propelled by its status as an important artificial intelligence technique. CA operates as an evolutionary system where cells, initially in a defined state, adhere to predefined rules governing interactions with their surroundings. Widely applicable, CA finds utility in simulating natural phenomena, robotic control, and modeling various systems across physical, biological, and sociological domains. This article presents a systematic review of literature on the application of CA in robotics, drawing insights from data generated through the StArt software. Our review, spanning a decade, aimed to identify relevant studies within the extant literature. Following rigorous evaluation, 31 articles were selected, covering themes such as path planning, surveillance, navigation, garbage collection, and foraging. However, our assessment uncovered a notable scarcity of research output in this domain, with only a limited number of authors contributing globally, many of whom are still in early stages of inquiry. Notably, a significant proportion of the selected works originated from Brazilian researchers, underscoring their pioneering contributions to advancing the field.

Prediction of Stock Prices Using Ensemble Models

Cláudio Estevam Leite da Silva — 2024-09-04

The financial market encompasses a set of institutions, products, and services aimed at meeting the financial needs of individuals, companies, and governments. Its primary objective is to direct financial resources from investors to projects requiring funding. This is achieved through the issuance and trading of securities such as stocks, debt securities, among others. In this paper, the goal was to develop a machine learning application specifically for the Brazilian financial market, focusing on predicting the market value of eight companies that are representative of the financial sector on the stock exchange. The prediction is based on the closing price history and uses data from the last three years, with the inputs corresponding to the last 60 days immediately preceding the forecast date. For this task, three machine learning models were selected: Long Short-Term Memory (LSTM), Multilayer Perceptron (MLP), and Convolutional Neural Network (CNN). Each of these was fine-tuned using five different optimizers, resulting in a total of 15 models. Subsequently, all 15 models were combined into an Ensemble. After applying data transformations, the models achieved a satisfactory level of error for the analysis. Among the transformations used, the logarithmic transformation stood out as the one that resulted in the most well-adjusted models compared to the others. In second place, the Yeo-Johnson transformation showed slightly higher error but performed better on series with high variation. Additionally, the convolutional models and Ensemble were the most effective.