Journal of Artificial Intelligence and Systems最新文献

{"title":"Classifying Alzheimer's disease based on a convolutional neural network with MRI images","authors":"M. Avşar, K. Polat","doi":"10.33969/ais.2023050104","DOIUrl":"https://doi.org/10.33969/ais.2023050104","url":null,"abstract":"Alzheimer's disease is a significant disease that negatively affects daily life and reduces the quality of human life. Dementia and Alzheimer's disease occur as the loss of neurons or a decrease in the relationship between neurons. So far, no effective drug has been found in diagnosing this disease. For this reason, it has become essential for individuals to diagnose the disease early and to detect the disease before it progresses. However, early diagnosis of the disease is challenging. The disease can be diagnosed after significant and irreversible effects on humans occur. A lot of research has been done worldwide for early disease diagnosis. Deep learning algorithms have become essential in diagnosing this disease. Significant progress has been made in diagnosing the disease with models created using deep learning algorithms. This study used a sequential model, conv2D, maxPooling2D, and dense layers to diagnose and classify. According to the dataset from Kaggle, a 4-class dataset has been used in this study to diagnose Alzheimer's disease. According to the Alzheimer's MRI dataset, the disease has been classified as nondemented, moderate demented, mild demented, and very mild demented, respectively. The proposed model has been trained using CNN. The number of layers and dropout rate have been used as performance metrics. In our study, activation Leaky ReLU was used. The SMOTE technique has been used to oversample the available data. This study's classification results will help experts make the right decisions. With F1Score, accuracy, recall, and precision values, 96.35% success was achieved in the CNN model. Different CNN methods can be used to advance these studies.","PeriodicalId":273028,"journal":{"name":"Journal of Artificial Intelligence and Systems","volume":"192 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124270104","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Feature Extraction aligned Email Classification based on Imperative Sentence Selection through Deep Learning","authors":"Nashit Ali, Anum Fatima, Hureeza Shahzadi, Aman Ullah, K. Polat","doi":"10.33969/ais.2021.31007","DOIUrl":"https://doi.org/10.33969/ais.2021.31007","url":null,"abstract":"Most commonly used channel for communication among peoples is emails. In this era where everyone is so busy in their routine and work, it is very difficult to check all email when one receives huge amount of emails. Previous research has done work on email categorization in which they have mostly done spam filtration. The problem with spam filtration is that sometimes person mistakenly mark an important email received from high authority as spam and according to previous research, this email will be filtered as spam that can cause a great threat for job of an employee. In this research, we are introducing a methodology which classifies email text into three categories i.e. order, request and general on basis of imperative sentences. This research use Word2Wec for words conversion into vector and use two approaches of deep learning i.e. Convolutional neural network and Recurrent neural network for email classification. We conduct experiment on Dataset collected from Personal Gmail account and Enron which consists of 1000 emails. The experiment result show that RNN gives better accuracy than CNN. We also compare our methods with previously used method Fuzzy ANN results and Our proposed methods CNN and RNN gives better results than Fuzzy ANN. This research has also included different experimental result in which CNN and RNN applied on different ratios of training and testing dataset. These experiment show that increasing in the ratio of training dataset results in increasing accuracy of algorithm.","PeriodicalId":273028,"journal":{"name":"Journal of Artificial Intelligence and Systems","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129718281","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Symbolic Regression Based Feature Extraction of Shallow Neural-Networks for Identification and Prediction","authors":"S. Beyhan","doi":"10.33969/ais.2022040103","DOIUrl":"https://doi.org/10.33969/ais.2022040103","url":null,"abstract":"This paper proposes a feature extraction method to improve the performance of shallow neural-network models with less number of parameters to apply especially on embedded system design at remote applications. Feature extraction method is designed using fuzzy c-means clustering based fuzzy system design cascaded a layer of symbolic operators and functions, respectively. During the training stage of neural-networks, symbolic operators and functions are selected using random-learning theory with the unity internal weights such that based on the prediction performance, optimal sequences are recorded for feature extraction to be utilized on testing phase. Extracted features are here used to empower the single-layer neural-network (SLNN) with sigmoid hyperbolic activation functions, functional-link neural- network (FLNN) with Chebyshev polynomials and Pi-Sigma higher-order neural-network (PSNN) with sigmoid activation functions, respectively. The internal and output parameters of the appended shallow neural-networks are optimized using batch optimization methods. Proposed regression models are first tested on identification of an artificial discrete-time dynamic system and real-time inverted pendulum then also for prediction of the sunspot time-series and traffic density estimation. As a result, the prediction performance of shallow neural networks is improved to be used in future applications.","PeriodicalId":273028,"journal":{"name":"Journal of Artificial Intelligence and Systems","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126165405","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Vision Sensor Assisted Fire Detection in IoT Environment using ConvNext","authors":"S. Zahir, W. Abbas, R. Khan, M. Ullah, Arbab Waseem, Rafi Ullah Abbas, M. Khan","doi":"10.33969/ais.2023050102","DOIUrl":"https://doi.org/10.33969/ais.2023050102","url":null,"abstract":"To mitigate social, ecological, and financial damage, effective fire detection and control are crucial. Performing real-time fire detection in Internet of Things (IoT) environments, however, presents significant challenges due to limited storage, transmission, and computational resources. Early fire detection and automated response are essential for addressing these challenges. In this paper, we introduce an IoT-supported deep learning model designed for efficient fire detection. The proposed model builds upon the pre-trained weights of the ConvNext convolutional neural network, which excels at detecting minute features and distinguishing between yellow lights and fire patterns. Implemented on an IoT device, the system triggers an alert when a fire is detected, prompting necessary actions. Our method, tested on the forest fire dataset, demonstrated a 4% improvement in accuracy compared to existing deep learning models for fire detection.","PeriodicalId":273028,"journal":{"name":"Journal of Artificial Intelligence and Systems","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124989311","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Performance Evaluation of Machine Learning Models for Weather Forecasting","authors":"Iliyas Ibrahim Iliyas, Andra Umoru, A. E. Chahari, Mustapha Mallam Garba","doi":"10.33969/ais.2022040102","DOIUrl":"https://doi.org/10.33969/ais.2022040102","url":null,"abstract":"Temperature is used to indicate variability and climate changes that indicate the process which is been carried out within the ecosystem and its services. The lack of knowledge about temperature affects human lives in terms of agriculture, transportation, mining, etc. temperature forecasting is used to predict atmospheric conditions based on parameters that caused the temperature to change. This study aims to explore the use of machine learning models for the prediction of temperature, evaluate the performance of these models, and use the model to predict temperature. In this study we explore the use of four different machine learning algorithms for forecasting weather temperature, the algorithms are: Ridge, Random Forest, Linear Regression, and Decision tree. We divided the dataset into training and testing sets, The models were tested on 1000 testing sets based on RMSE score with Decision Tree having the best score of 0.036, Random Forest: 0.208 while Logistic Regression and Ridge had the lowest score of 0.759 respectively.","PeriodicalId":273028,"journal":{"name":"Journal of Artificial Intelligence and Systems","volume":"73 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125983878","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Trajectories modelling of mesoscale anticyclonic eddies in the Mozambique Channel using ANFIS Fuzzy C-Means","authors":"Hanitra Elisa Rasoavololoniaina, Harimino Andriamalala Rajaonarisoa, Roselin Randrianantenaina, A. Ratiarison, Hanitra Elisa, Rasoavololoniaina, Harimino Andriamalala, Rajaonarisoa, Todihasina Roselin, Adolphe Randrianantenaina, Andriamanga","doi":"10.33969/ais.2023050105","DOIUrl":"https://doi.org/10.33969/ais.2023050105","url":null,"abstract":"The aim of this paper is to optimize the Fuzzy C-Means (FCM) model of the ANFIS neuro-fuzzy system to model the four types of mesoscale anticyclonic eddy trajectories in the Mozambique Channel as a function of the variables eddy speed average of contour, amplitude and diameter, horizontal wind, atmospheric pressure and bathymetry. The study area concerns the eastern part of the Mozambique Channel between longitudes 41°E-44°E and latitudes 16°S-25°S. We classified the eddy trajectories of interest in our study area into four types according to their formation and dissipation zones. The data used are from the mesoscale eddy track atlas product derived from the META3 altimetry version. 1exp DT allsat for trajectories and eddy properties (amplitude, eddy rotation speed and diameter), GEBCO_2022 grid data for bathymetry, ECMWF data at spatial resolution 1° x 1° for atmospheric pressure, and Copernicus Marine data at spatial resolution 0.25° x 0.25° for wind. The latitudes and longitudes of the daily eddy displacement points from their formation to their dissipation characterize the trajectories. We used two different approaches in our study. The first approach consist to put each endogenous variable as input for the FCM model, while the second approach utilized the endogenous variables multiplied by the multiple regression coefficients. The results conclude that the case where the input variables of the model are preprocessed by the multiple (linear or polynomial) regression operation before FCM modeling is the best approach.","PeriodicalId":273028,"journal":{"name":"Journal of Artificial Intelligence and Systems","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124375747","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An Overview of Deep Learning Models for Foliar Disease Detection in Maize Crop","authors":"Jagrati Paliwal, S. Joshi","doi":"10.33969/ais.2022040101","DOIUrl":"https://doi.org/10.33969/ais.2022040101","url":null,"abstract":"Agriculture is an important sector of Indian economy and India is among the top three global producers of agricultural products. Protecting the crops and producing healthy yields is a prime goal of the agriculture industries. The agricultural crops are susceptible to diseases and demands proactive early diagnosis and treatment. Studies and Research are in progress to find smart methods and techniques for accurate diagnosis of crop diseases to prevent major yield losses and financial losses. The present study outlines the role of Deep Learning in the crop disease detection and discusses the future advancements in maize disease detection. The paper focuses on the role of Deep Learning in identification of diseases on maize plant leaf and describes about some common maize diseases and its classification methods. The paper shall help readers to gain insight on Deep Learning techniques to solve classification problems and encourage them to proceed for future work in the concerned domain.","PeriodicalId":273028,"journal":{"name":"Journal of Artificial Intelligence and Systems","volume":"92 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126588684","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Robotic Inspection Monitoring System for Pipelines","authors":"Muhammad Ahmad Baballe","doi":"10.33969/ais.2022040104","DOIUrl":"https://doi.org/10.33969/ais.2022040104","url":null,"abstract":"The most popular method for transporting fluids and gases is through pipelines nowadays. Regular inspection is necessary for the pipelines to work correctly. Humans must not enter potentially dangerous environments to inspect these pipelines. As a result of this, pipeline robots came into existence. These pipe inspection robots help in pipeline inspection, protecting numerous people from harm since human beings cannot enter the pipes and inspect them in case there is any such or kind of damage that requires repair. Despite numerous improvements, pipeline robots still have several limitations. The introduction of this in pipe inspection robots helps to solve many problems, such as leakage of the gas or fluid pipelines, rustiness, and also if the pipe is broken from any part.","PeriodicalId":273028,"journal":{"name":"Journal of Artificial Intelligence and Systems","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116917666","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Machine Learning Methods for Predicting the Popularity of Movies","authors":"D. Oyewola, E. Dada","doi":"10.33969/ais.2022040105","DOIUrl":"https://doi.org/10.33969/ais.2022040105","url":null,"abstract":"The movie industry has grown into a several billion-dollar enterprise, and there is now a ton of information online about it. Numerous machine learning techniques have been created by academics and can produce effective classification models. In this study, different machine learning classification techniques are applied to our own movie dataset for multiclass classification. This paper's main objective is to compare the effectiveness of various machine learning techniques. This study examined five methods: Multinomial Logistic Regression (MLR), Support Vector Machine (SVM), Bagging (BAG), Naive Bayes (NBS) and K-Nearest Neighbor (KNN), while noise was removed using All K-Edited Nearest Neighbors (AENN). These techniques all utilize previous IMDb dataset to predict a movie's net profit value. The algorithms predict the profit at the box office for each of these five techniques. Based on the dataset used in this paper, which consists of 5043 rows and 14 columns of movies, this study evaluates the performance of all seven machine learning techniques. Bagging outperformed other machine learning techniques with a 99.56% accuracy rate.","PeriodicalId":273028,"journal":{"name":"Journal of Artificial Intelligence and Systems","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126286554","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Neural Network Learning of Context-Dependent Affordances","authors":"Luca Simione, A. Borghi, S. Nolfi","doi":"10.33969/ais.2022040106","DOIUrl":"https://doi.org/10.33969/ais.2022040106","url":null,"abstract":"In this paper, we investigated whether affordances are activated automatically, independently of the context in which they are experienced, or not. The first hypothesis postulates that stimuli affording different actions in different contexts tend to activate all actions initially. The action appropriate to the current context is later selected through a competitive process. The second hypothesis instead postulates that only the action appropriate to the current context is activated. The apparent tension between these two alternative hypotheses constitutes an open issue since, in some cases, experimental evidence supports the context-independent hypothesis, while in other cases it supports the context-dependent hypothesis. To study this issue, we trained a deep neural network with stimuli in which action inputs co-varied systematically with visual inputs. The neural network included two separate pathways for encoding visual and action inputs with two hidden layers each, and then a common hidden layer. The training was realized through an auto-associative unsupervised learning algorithm and the testing was conducted by presenting only part of the stimulus to the neural network, to study its generative properties. As a result of the training process, the network formed visual-action affordances. Furthermore, we conducted the training process in different contexts in which the relation between stimuli and actions varied. The analysis of the obtained results indicates that the network displays both a context-dependent activation of affordances (i.e., the action appropriate to the current context tends to be more activated than the alternative action) and a competitive process that refines action selection (i.e., that increases the offset between the activation of the appropriate and unappropriate actions). Overall, this suggests that the apparent contradiction between the two hypotheses can be resolved. Moreover, our analysis indicates that the greater facility with which colour-action associations are acquired with respect to shape-action associations is because the representation of surface features, such as colour, tends to be more readily available for deeper features, such as shape. Our results support the feasibility of human-like affordance acquisition in artificial neural networks trained using a deep learning algorithm. This model could be further applied to a number of robotic and applicative scenarios.","PeriodicalId":273028,"journal":{"name":"Journal of Artificial Intelligence and Systems","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134062067","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}