Inteligencia Artificial <p style="text-align: justify;"><span style="color: #000000;"><strong><em><a style="color: #003366; text-decoration: underline;" href="" target="_blank" rel="noopener">Inteligencia Artificial</a></em></strong><span id="result_box" class="" lang="en"> is an international open access journal promoted by <span class="">the Iberoamerican Society of</span> Artificial Intelligence (<a href="">IBERAMIA</a>). </span></span>Since 1997, the journal publishes high-quality original papers reporting theoretical or applied advances in all areas of Artificial Intelligence. <span style="color: rgba(0, 0, 0, 0.87); font-family: 'Noto Sans', -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Oxygen-Sans, Ubuntu, Cantarell, 'Helvetica Neue', sans-serif; font-size: 14px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: justify; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; background-color: #ffffff; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial; display: inline !important; float: none;">There are no fees for subscription, publication nor editing tasks<span class="VIiyi" lang="en"><span class="JLqJ4b ChMk0b" data-language-for-alternatives="en" data-language-to-translate-into="es" data-phrase-index="0">.</span></span> <span class="VIiyi" lang="en"><span class="JLqJ4b ChMk0b" data-language-for-alternatives="en" data-language-to-translate-into="es" data-phrase-index="0">Articles can be written in English, Spanish or Portuguese and <a href="">will be subjected</a> to a double-blind peer review process.</span></span> <span class="VIiyi" lang="en"><span class="JLqJ4b ChMk0b" data-language-for-alternatives="en" data-language-to-translate-into="es" data-phrase-index="0">The journal is abstracted and indexed in several <a href="">data bases</a>.</span></span><br /></span></p> Sociedad Iberoamericana de Inteligencia Artificial (IBERAMIA) en-US Inteligencia Artificial 1137-3601 <p>Open Access publishing.<br />Lic. under <a href="">Creative Commons CC-BY-NC</a><br />Inteligencia Artificial (Ed. IBERAMIA)<br />ISSN: 1988-3064 (on line).<br />(C) IBERAMIA &amp; The Authors</p> Evaluating the impact of curriculum learning on the training process for an intelligent agent in a video game <p>We want to measure the impact of the curriculum learning technique on a reinforcement training setup, several experiments were designed with different training curriculums adapted for the video game chosen as a case study. Then all were executed on a selected game simulation platform, using two reinforcement learning algorithms, and using the mean cumulative reward as a performance measure. Results suggest that curriculum learning has a significant impact on the training process, increasing training times in some cases, and decreasing them up to 40% percent in some other cases.</p> Jorge E Camargo Rigoberto Sáenz Copyright (c) 2021 Iberamia & The Authors 2021-09-30 2021-09-30 24 68 1 20 10.4114/intartif.vol24iss68pp1-20 A New Method of Different Neural Network Depth and Feature Map Size on Remote Sensing Small Target Detection <p>Convolutional neural networks (CNNs) have shown strong learning capabilities in computer vision tasks such as classification and detection. Especially with the introduction of excellent detection models such as YOLO (V1, V2 and V3) and Faster R-CNN, CNNs have greatly improved detection efficiency and accuracy. However, due to the special angle of view, small size, few features, and complicated background, CNNs that performs well in the ground perspective dataset, fails to reach a good detection accuracy in the remote sensing image dataset. To this end, based on the YOLO V3 model, we used feature maps of different depths as detection outputs to explore the reasons for the poor detection rate of small targets in remote sensing images by deep neural networks. We also analyzed the effect of neural network depth on small target detection, and found that the excessive deep semantic information of neural network has little effect on small target detection. Finally, the verification on the VEDAI dataset shows, that the fusion of shallow feature maps with precise location information and deep feature maps with rich semantics in the CNNs can effectively improve the accuracy of small target detection in remote sensing images.</p> Yaming Cao ZHEN YANG CHEN GAO Copyright (c) 2021 Iberamia & The Authors 2021-09-30 2021-09-30 24 68 21 32 10.4114/intartif.vol24iss68pp21-32 Procedural Content Generation for General Video Game Level Generation <p>With the passage of time, video games are becoming more complex, and their development incurs greater<br>time and cost. The creation of video gaming content such as levels, maps, textures and so on represent a large<br>part of the overall cost of game development. Procedural Content Generation (PCG) is a method of generating<br>content via a pseudo-random process. Level generation has been the most signi cant and oldest problem in the<br>PCG domain. The majority of the PCG level generators are speci c to a particular game, content is generated<br>only for a suited single type and these generators are evaluated mostly by computational metrics, user studies<br>and tness functions. Considering, the grand goal of general Arti cial Intelligence, it would be bene cial to sculpt<br>solutions that are applicable to a general set of problems. For the level generation problem, this can be achieved<br>by constructing a level generator that generates levels for a set of games and not explicitly for a single game. In<br>this research, we have created four di erent type of generators for the GVG-LG framework. The generators follow<br>a distinct path and are able to solve multiple problems related to PCG including dynamic diculty adjustment,<br>creation of intelligent controllers, creating aesthetically appealing levels and using patterns as objectives for level<br>generation. In addition, we evaluated all the generators using a variety of techniques. The experimental results<br>show promising results and represent our attempt at general video game level generation.</p> Adeel Zafar Hasan Mujtaba Omer Beg Copyright (c) 2021 Iberamia & The Authors 2021-09-30 2021-09-30 24 68 33 36 10.4114/intartif.vol24iss68pp33-36 KeyFinder: An Efficient Minimal Keys Finding Algorithm For Relational Databases <p>In relational databases, it is essential to know all minimal keys since the concept of database normaliza-<br>tion is based on keys and functional dependencies of a relation schema. Existing algorithms for determining keys<br>or computing the closure of arbitrary sets of attributes are generally time-consuming. In this paper we present an<br>efficient algorithm, called KeyFinder, for solving the key-finding problem. We also propose a more direct method<br>for computing the closure of a set of attributes. KeyFinder is based on a powerful proof procedure for finding<br>keys called tableaux. Experimental results show that KeyFinder outperforms its predecessors in terms of search<br>space and execution time.</p> Moussa Demba Copyright (c) 2021 Iberamia & The Authors 2021-09-30 2021-09-30 24 68 37 52 10.4114/intartif.vol24iss68pp37-52 Applying ensemble neural networks to analyze industrial maintenance: Influence of Saharan dust transport on gas turbine axial compressor fouling <p>The planning of industrial maintenance associated with the production of electricity is vital, as it yields a current and future snapshot of an industrial component in order to optimize the human, technical and economic resources of the installation. This study focuses on the degradation due to fouling of a gas turbine in the Canary Islands, and analyzes fouling levels over time based on the operating regime and local meteorological variables. In particular, we study the relationship between degradation and the suspended dust that originates in the Sahara Desert. To this end, we use a computational procedure that relies on a set of artificial neural networks to build an ensemble, using a cross-validated committees approach, to yield the compressor efficiency. The use of trained models makes it possible to know in advance how the local fouling of an industrial rotating component will evolve, which is useful for maintenance planning and for calculating the relative importance of the variables that make up the system</p> D. Gonzalez-Calvo R.M. Aguilar C. Criado-Hernandez L.A. Gonzalez-Mendoza Copyright (c) 2021 Iberamia & The Authors 2021-09-30 2021-09-30 24 68 53 71 10.4114/intartif.vol24iss68pp53-71 The Effect of the Dataset Size on the Accuracy of Software Defect Prediction Models: An Empirical Study <p><strong>The ongoing development of computer systems requires massive software projects. Running the components of these huge projects for testing purposes might be a costly process; therefore, parameter estimation can be used instead. Software defect prediction models are crucial for software quality assurance. This study investigates the impact of dataset size and feature selection algorithms on software defect prediction models. We use two approaches to build software defect prediction models: a statistical approach and a machine learning approach with support vector machines (SVMs). The fault prediction model was built based on four datasets of different sizes. Additionally, four feature selection algorithms were used. We found that applying</strong><strong> the SVM defect prediction model on datasets with a reduced number of measures as features may enhance the accuracy of the fault prediction model. Also, it directs the test effort to maintain the most influential set of metrics. We also found that the running time of the SVM fault prediction model is not consistent with dataset size. Therefore, having fewer metrics does not guarantee a shorter execution time. From the experiments, we found that dataset size has a direct influence on the SVM fault prediction model. However, reduced datasets performed the same or slightly lower than the original datasets.</strong></p> Mashaan A. Alshammari Mohammad Alshayeb Copyright (c) 2021 Iberamia & The Authors 2021-10-26 2021-10-26 24 68 72 88 10.4114/intartif.vol24iss68pp72-88 Performance Analysis in the Segmentation of urban asphalted roads in RGB satellite images using K-Means++ and SegNet <p>The design and manual insertion of new terrestrial roads into geographic databases is a frequent activity in geoprocessing and their demand usually occurs as the most up-to-date satellite imagery of the territory is acquired. Continually, new urban and rural occupations emerge, for which specific vector geometries need to be designed to characterize the cartographic inputs and accommodate the relevant associated data. Therefore, it is convenient to develop a computational tool that, with the help of artificial intelligence, automates what is possible in this respect, since manual editing depends on the limits of user agility, and does it in images that are usually easy and free to access.</p> <p>To test the feasibility of this proposal, a database of RGB images containing asphalted urban roads is presented to the K-Means++ algorithm and the SegNet Convolutional Neural Network, and the performance of each was evaluated and compared for accuracy and IoU of road identification.</p> <p>Under the conditions of the experiment, K-Means++ achieved poor and unviable results for use in a real-life application involving tarmac detection in RGB satellite images, with average accuracy ranging from 41.67% to 64.19% and average IoU of 12.30% to 16.16%, depending on the preprocessing strategy used. On the other hand, the SegNet Convolutional Neural Network proved to be appropriate for precision applications not sensitive to discontinuities, achieving an average accuracy of 87.12% and an average IoU of 71.93%.</p> João Batista Pacheco Junior Henrique Mariano Costa do Amaral Copyright (c) 2021 Iberamia & The Authors 2021-11-24 2021-11-24 24 68 89 103 10.4114/intartif.vol24iss68pp89-103 An Accurate Integrated System to detect Pulmonary and Extra Pulmonary Tuberculosis using Machine Learning Algorithms <p>Several studies have been reported the use of machine learning algorithms in the detection of Tuberculosis, but studies that discuss the detection of both types of TB, i.e., Pulmonary and Extra Pulmonary Tuberculosis, using machine learning algorithms are lacking. Therefore, an integrated system based on machine learning models has been proposed in this paper to assist doctors and radiologists in interpreting patients’ data to detect of PTB and EPTB. Three basic machine learning algorithms, Decision Tree, Naïve Bayes, SVM, have been used to predict and compare their performance. The clinical data and the image data are used as input to the models and these datasets have been collected from various hospitals of Jalandhar, Punjab, India. The dataset used to train the model comprises 200 patients’ data containing 90 PTB patients, 67 EPTB patients, and 43 patients having NO TB. The validation dataset contains 49 patients, which exhibited the best accuracy of 95% for classifying PTB and EPTB using Decision Tree, a machine learning algorithm.</p> Rupinder Kaur Anurag Sharma Copyright (c) 2021 Iberamia & The Authors 2021-12-21 2021-12-21 24 68 104 122 10.4114/intartif.vol24iss68pp104-122 Greedy Genetic Algorithm for the Data Aggregator Positioning Problem in Smart Grids <p>In this work, we present a metaheuristic based on the genetic and greedy algorithms to solve an application of the set covering problem (SCP), the data aggregator positioning in smart grids. The GGH (Greedy Genetic Hybrid) is structured as a genetic algorithm, but it has many modifications compared to the classic version. At the mutation step, only columns included in the solution can suffer mutation and be removed. At the recombination step, only columns from the parent’s solutions are available to generate the offspring. Moreover, the greedy algorithm generates the initial population, reconstructs solutions after mutation, and generates new solutions from the recombination step. Computational results using OR-Library problems showed that the GGH reached optimal solutions for 40 instances in a total of 75 and, in the other instances, obtained good and promising values, presenting a medium gap of 1,761%.</p> Sami Nasser Lauar Mario Mestria Copyright (c) 2021 Iberamia & The Authors 2021-12-21 2021-12-21 24 68 123 137 10.4114/intartif.vol24iss68pp123-137 Symmetry-Based Brain Abnormality Detection Using Machine Learning <p>Medical image processing, which includes many applications such as magnetic resonance image (MRI) processing, is one of the most significant fields of computer-aided diagnostic (CAD) systems. the detection and identification of abnormalities in the magnetic resonance imaging of the brain is one of the important applications that uses magnetic resonance imaging and digital image processing techniques. In this study, we present a method that relies on the symmetry and similarity between the two lobes of the brain to determine if there are any abnormalities in the brain because tumours cause deformations in the shape of one of the lobes, which affects this symmetry. The proposed approach overcomes the challenge arising from different shapes of brain images of different people, which poses an obstacle to some approaches that rely on comparing one person’s brain image with other people's brain images. In the proposed method the image of the brain is divided into two parts, one for the left lobe and the other for the right lobe. Some measures are extracted from the features of the image of each lobe separately and the distance between the corresponding metrics are calculated. These distances are used as the independent variables of the classification algorithm which determines the class to which the brain belongs. Metrics extracted from various features, such as colour and texture, were studied, discussed and used in the classification process. The proposed algorithm was applied to 366 images from standard datasets and four classifiers were tested namely Naïve Bayes (NB), random forest (RF), logistic regression (LR), and support vector machine (SVM). The obtained results from these classifiers have been discussed thoroughly and it was found that the best results were obtained from RF classifiers where the accuracy was 98.2%. Finally, The results obtained and the limitations were discussed and benchmarked with state-of-the-art approaches.</p> Mohammad Al-Azawi Copyright (c) 2022 Iberamia & The Authors 2022-01-19 2022-01-19 24 68 138 150 10.4114/intartif.vol24iss68pp138-150