International Journal of Advanced Intelligence Paradigms (17 papers in press)

Regular Issues

• Big data secure storing in cloud and privacy preserving mechanism for outsourced cloud data  
  by Dr B. Renuka 
  Abstract: Big data is a buzz word in this decade it gets tremendous concentration in these days by the researchers because of the characteristics and features. And also big data gives lot of challenges to the world that is storage, processing and security. In any technology security is the prime concern in this manuscript, we map to misuse new complications of enormous information regarding security, further more, confer our thought toward viable and insurance protecting enlisting in the immense data time. Specifically, we at first formalize the general building of gigantic data examination, recognize the relating security necessities, and present a capable and assurance sparing outline for immense data which is secured in cloud.
  Keywords: Privacy Preserving; Security; Big data; Cloud Computing; outsourced data.
  
  
• A Survey on Wireless Networks to Balancing the Load in Wireless Mesh  
  by Subba Rao 
  Abstract: Now a days wireless technology occupy the very prominent role in all the sectors and WMNs have play the prominent role in coming generations because it has many benefits over other wireless networks. However, still there are many technical issues are there which are going to discussed in WMN. The major problem of WMNs is sustained the balancing. In WMNs, comparison between the incoming data traffic to a node is greater than the outgoing data traffic, then congestion is high in this network. Various authors have proposed to reduce the congestion and for improving the network throughput. This paper discussed about analysis of various load balancing techniques to facilitate the researchers as well as practitioners in choosing a proper load balancing technique for improving the network performance.
  Keywords: Wireless mesh network;Gateway;Load balance;Path;Router;NS2; etc.
  
  
• Prediction and Detection of Kidney Diseases using Ensemble Classification  
  by Suresh Babu 
  Abstract: Chronic Kidney Disease (CKD) is the most dangerous disease occurred in many of the persons now days. As the research is going on there is exact reason for damaging of kidneys. Some research says that this may be due to the lack of drinking water and high blood pressure and Diabetes and some of the diseases such as severe dehydration, kidney trauma. After all the research done, in this paper, the new prediction and detection of chronic kidney disease using ensemble data mining classification is described for the better results.
  Keywords: CKD; Diabetes; Diabetes.
  
  
• Towards Smart Healthcare System in Airlines with IoT and Cloud Computing  
  by Veeraa Anjaneyuluu 
  Abstract: The world is becoming smart with the advances brought in the new technologies. In the recent past, Internet of Things(IoT) is playing major role in all the fields of the world. Smart health care application are developing with many accepts and Cloud Computing also effective part for data communication around the world. Detection and controlling of contagious diseases is also a major issue, when people travel all over the world in airways. In this paper we propose architecture to smart identification of the person/s with diseases while travelling in airlines. Adoption of this architecture able to control and stop the pervasive healthcare. An effective ways are discussed and to determine the percentage of infection of particular disease using probability tables. Based on our concept and results we are also given directions to development of tools and applications.
  Keywords: Internet of Things; Cloud Computing; Contagious Diseases; Airlines System.
  
  
• MULTI-MODAL FEATURE FUSION AND ITS APPLICATION TO BRAIN TUMOR CLASSIFICATION  
  by Jyothsna Devi 
  Abstract: Tumor is a mass of abnormal cells. The tumor that grows inside human skull is termed as brain tumor. Human brain is enclosed by the skull. Tumor that grows in such restricted space can cause problems. Tumors that grow in the brain are categorized as cancerous or noncancerous. If these types of tumors cannot be detected at their early stages may lead to brain damage, and it can be life-threatening. In this work we have used SVM to detect whether the given MRI image of a brain tumor is malignant or benign. Recent literature shows that Support Vector Machines (SVM) is a supervised classification technique that has increased popularity as they exhibit high generalization ability even trained with small set of training data. SVM has good generalization ability to solve many real-time problems. In this work we have used SVM based classifier to identify whether the tumor is malignant or benign. Initially hand crafted features like Discrete Wavelet Transform (DWT) or gist features are extracted from the given MR images, then follows preprocessing and segmentation tasks followed by SVM based classification. Each of these representations have their own advantages of representing images. If we consider any single representation then we are ignoring the advantages of using the other representation. Our proposed method tried to exploit the benefit from different representations of the images. Motivated from the fusion based classification models, in this work we have extracted different representations from the given MR images and fused them to represent the image as a single feature vector. We have applied different fusion techniques to improve the performance of the SVM based tumor classification. Our experimental studies on bench mark datasets show that fusion techniques can enhance the accuracy of SVM classification for brain tumor classification. Along with fusion we have also tried to examine the efficiency of various kernels on the classifiers performance.
  Keywords: Support Vector Machine; Otsu Segmentation; Discrete Wavelet Transform; Non local mean filter; fusion.
  
  
• CUSTOMER REVIEW SUMMARIZATION BASED ON GENERIC WRAPPER AND OPINION MINING  
  by K. Priya, K. Dinakaran 
  Abstract: Now-a-days onine shopping by customers is getting increased day by day. Customers are having awareness in the case of buying products based on the features of the products. The features of the product may be model, colour, size, Durability or price.The customer reviews or feedbacks based on the price of the products are collected from three different shopping websites and then consolidated and also ranked under separate website.The customer will be buying the product based on lowest price ,which online shopping website is holding.Through this work customer can avoid confusion while shopping for products.Generic wrapper and VIPS techniques are used.These details can also be posted or shared in social networks webpages for customers convenience.Then the customers can maintain budjet and preventing them from taking wrong decision during online shopping.Searching Time for product information under each website can be reduced for the customers.
  Keywords: Generic wrapper; Wrapper Generation; product review; social network; Review Summarization; E-Commerce.
  DOI: 10.1504/IJAIP.2021.10030025
   
  
• Combination of Machine Learning methods to Solve Cold start problem in Recommender system  
  by Nitin Mishra, Vimal Mishra 
  Abstract: Recommender systems are special type of intelligent systems which exploits historical of user rating on items to make recommendation of items for those users. They are used in wide range of applications like online shopping, E-Commerce services social networking applications and many more. These are also used in banks and other services. They can also be used in fault finding in critical systems. In our paper we are discussing a problem known as cold start problem where you new user has a problem as he has missing history. We have used clustering approach to cluster users and then using these cluster labels for supervised machine learning to solve the cold start problem of new users. We have validated our solution on movielens dataset and found it to be solving cold start problem in a magical way. so we are claiming our approach to be a novel approach for solving cold start problem using combination of several methods some of which belong to collaborative filtering domain and others belong to content based domain. We have done exhaustive check so that it could be our fault free solution. As our method predict certain items so the results are accurate to our domain and can vary with small amount in similar domains. But theoretically, our Method can be used for solving cold start problem in general in any domain.We also claim that our method performance becomes better with increasing value of N of TopN Recommendation.
  Keywords: Cold start problem;Recommender systems; Machine learning;classification;clustering;k-modes clustering.
  
  
• A Novel Variant of Bat Algorithm Inspired from CATD-Pursuit Strategy & Its Performance Evaluations  
  by Shabnam Sharma, Sahil Verma, Kiran Jyoti 
  Abstract: This paper presents a novel nature inspired optimization technique, which is a variant of Standard Bat Algorithm. This optimization technique is inspired from the pursuit strategy of microchiroptera bats and their efficient way of adaptation according to dynamic environment. Here dynamic environment describes different movement strategies adopted by prey (target), during their pursuit. Accordingly, bats have to adopt different pursuit strategies to capture the prey (target). In this research work, a variant of Bat Algorithm is proposed considering the pursuit strategy Constant Absolute Target Detection (CATD), adopted by bats, while targeting preys moving erratically. The proposed algorithm is implemented in Matlab. Results obtained are validated in comparison to Standard Bat Algorithm on the basis of best, mean, median, worst and standard deviation. The results demonstrate that the proposed algorithm provides better exploration and avoid trapping in local optimal solution.
  Keywords: Bat Algorithm; Constant Absolute Target Detection (CATD); Computational Intelligence; Echolocation; Meta-heuristic; Nature-Inspired Intelligence; Optimization; Pursuit Strategy; Swarm Intelligence.
  DOI: 10.1504/IJAIP.2021.10030248
   
  
• Wireless Smart Automation Using IOT Based Raspberry Pi  
  by Vasu Goel, Akash Deep, Madireddy Vivek Reddy, Yedukondala Rao Veeranki 
  Abstract: In this paper we propose a smart door lock system and lighting system for home automation. This door lock system and lighting system is controlled by Radio Frequency Identification (RFID) reader which is programmed by Raspberry Pi to detect the input swipe through our university combo card or a RFID tag and wirelessly sends the signal to the Espruino (ESP) Wi-Fi module and Node Microcontroller Unit (MCU) which in turn activates the lighting system and door lock system. The mainstream application of the system will be in hostel rooms or in our homes wherever door locks are there so that doors can be opened anytime we want without disrupting our work or getting up from our places in case of any injury with a swipe of card
  Keywords: Internet-Of-Things; Raspberry pi; Radio-Frequency Identification; Home automation; MQTT.
  DOI: 10.1504/IJAIP.2019.10026853
   
  
• Data Mining Techniques and Fuzzy Logic to Build a Risk Prediction System for Stroke  
  by Farzana Islam, M. Rashedur Rahman 
  Abstract: Nowadays, by using different computational system medical sector predict diseases. These systems not only aid medical experts but also normal people. In recent years stroke becomes life threatening deadly cause and it increased at global alarming state. Early detection of stroke disease can be helpful to make decision and to change the lifestyle of people who are at high risk. There is a high demand to use computational expertise for prognosis stroke. Research has been attempted to make early prediction of stroke by using data mining techniques. This paper proposes rule based classifier along with other techniques. The dataset is collected from Dhaka medical college, situated in Dhaka, Bangladesh To build a more accurate and acceptable model the system uses different classification methods likely- Decision tree, Support vector machine, Artificial neural network and fuzzy model. K-means, EM and fuzzy C-means clustering algorithm are used to label the dataset more accurately. Fuzzy inference system is also built to generate rules. ANFIS provides the most accurate model.
  Keywords: stroke; decision tree; SVM; MLP; artificial neural network; support vector machine; fuzzy model; FIS; ANFIS; data-mining; fcm; clustering; EM clustering; k-means; Bangladeshi dataset; fuzzy rule.
  DOI: 10.1504/IJAIP.2021.10054275
   
  
• An optimized fuzzy edge detector for image processing and their use in modular neural networks for pattern recognition  
  by Isidra Espinosa-Velazquez, Patricia Melin, Claudia Gonzalez, Frumen Olivas 
  Abstract: In this paper, the development of a fuzzy edge detector optimized with the metaheuristics: Genetic Algorithms and Particle Swarm Optimization is presented, based on the sum of differences method, using as inputs the absolute values of the difference from the pixels in the image. The Pratts figure of merit metric was used to know the performance of the proposed fuzzy edge detector. A modular neural network was designed for the recognition of faces in benchmark images and comparisons were made with different works carried out with other fuzzy edge detection systems. The main contribution of this research work is the development of a new fuzzy edge detector method optimized.
  Keywords: fuzzy logic; fuzzy edge detector; optimization; GA; genetic algorithm; PSO; Particle swarm Optimization; Neural networks.
  
  
• PEBD: Performance Energy Balanced Duplication Algorithm for Cloud Computing  
  by Sharon Priya Surendran, Aisha Banu W 
  Abstract: With the increasing demand of cloud data, efficient task scheduling algorithms are required with minimal power consumption. In this paper, the Performance-Energy Balanced Duplication (PEBD) scheduling approach is proposed for energy conservation at the point of task duplication. Initially, the resources are preprocessed with the Manhattan distance based Fuzzy Clustering (MFC).Then resources are scheduled using a Novel duplication aware fault tolerant based League-BAT algorithm and faults expected during job executions can be handled proactively. The fault adaptive firefly optimization is used for minimizing faults and it keeps information about resource failure. Consequently, the optimization ensures that performance is improved with the help of task duplication with low energy consumption. The duplications are restricted and they are strictly forbidden if they provide significant enhancement of energy consumption. Finally, enhanced compress & Join algorithm is used for efficient compression processing. It considers both schedule lengths and energy savings to enhance the scheduling performance with less power consumption. The performance of energy consumption and makespan for the proposed approach is increased with 6% and 0.5 % respectively
  Keywords: Manhattan distance; Fuzzy clustering; Resource scheduling; Duplication; fault tolerance; energy conservation.
  
  
• Load-balanced Multilayered Clustering Protocol to Maximize the Lifetime of Wireless Sensor Networks  
  by Rohan Gupta, Arnab Nandi 
  Abstract: This article introduces an innovative clustering protocol for load balancing in Wireless Sensor Networks (WSNs). In the proposed protocol, square shape clusters of equal area are arranged in a multilayer fashion, and the base station is at the center of the network. The equal area of square clusters offers a nearly equal number of member nodes in each cluster which leads to comparable energy consumption at cluster heads for transmitting and receiving data from member nodes. This article also introduces a new routing approach in which hop selection is based on the difference of angle between the source and destination cluster heads with respect to a particular point. The efficiency of the proposed protocol concerning network lifetime and energy consumption is evaluated and compared with Low-Energy Adaptive Clustering Hierarchy (LEACH), Enhanced-Modified LEACH (E-MODLEACH) and Least Distance clustering (LDC). The efficiency of the proposed protocol is also evaluated for different optimization algorithms like GWO, PSO, and GSA. The proposed protocol is implemented with these algorithms during the cluster formation stage.
  Keywords: WSN; Clustering Protocol; Load Balancing; Network Lifetime; GWO; PSO; GSA; LEACH; E-MODLEACH; LDC.
  
  
• Case-Based Reasoning Methodology for eLearning Recommendation System  
  by Swati Shekapure, Dipti D. Patil 
  Abstract: Increasingly, eLearning has become a leading development trend in the industry. As far as the learning methodology is concerned, it has been observed that traditional learning methods such as teacher and student, chalk and duster have turned to modern & innovative learning. Due to a revolution in technology, everyone started learning by using the internet. They have been using devices like smartphones, laptops, e-books, I-pod and so on for gaining instructions. So, while they procure the learning they admit certain records, which are not significant to answer all their exploratory questions. Ultimately, there was a huge delay while scrutinizing the essential material on the internet, so there was an extremity to customize the search by acquiring certain information of a user to improve the search quality and save time. The recommended eLearning system is a case based system using a case-based reasoning approach and a distinct classification algorithmic rule to categorize the students learning interest. This system assembles student's learning preferences from a distinct discussion and systematically categorizes that characteristic into a learning standard.
  Keywords: Case-Based Reasoning; K Nearest Neighbor; Learning Style; Recommendation system.
  DOI: 10.1504/IJAIP.2022.10035296
   
  
• A Community Based Trusted Collaborative Filtering Recommender Systems Using Pareto Dominance Approach  
  by Anupama Angadi, Satya Keerthi Gorripati 
  Abstract: Recommender System algorithms provided clarification to information overload problem suffered by netizens. The Collaborative Recommender Filtering approach takes the user-item rating matrix as an input and recommends items based on the perceptions of similar neighbours. However, sparsity issue in the rating matrix leads to untrustworthy predictions. However, the conventional Collaborative Recommender Filtering method chooses ineffective descriptive users as neighbours for each target user. This hints that the recommendations made by the system remain inaccurate. The proposed approach addresses this issue by applying a pre-filtering process and integrates community detection with Pareto dominance, which considers trusted neighbours from the community into which the active user pertains and eliminates dominant users from the neighbourhood. The results on the proposed framework showed a noteworthy improvement in all the accuracy measures when related to the traditional approaches.
  Keywords: Community Detection; Recommender Systems; Sparsity; Pareto dominance; Cold Start; Trust propagation;.
  
  
• DOCUMENT SUMMARIZATION USING RECURRENT NEURAL NETWORK  
  by Vijayakumar K, Dafni Rose J 
  Abstract: Document Summarization is the process which condenses the given document to generate a summary which captures the main essence of the entire document. In recent years, there has been increased interest in automatic summarization. Automatic summarization refers to summarizing a document using software and it helps to reduce large text documents to a short set of words or a paragraph that delivers the main meaning of the full text. The extracted features from the documents are used for the automatic summarization process and remain a successfully proven approach but it leads to drawbacks with respect to structure, redundancy, coherence. Existing methods for single document summarization usually make use of only the first sentence or fixed number of words from the beginning contained in the specified document. This paper proposes a technique that uses contents of the entire document to provide more knowledge to help single document summarization. The proposed system mainly aims at generating a summary of at least a minimum length unlike the existing system that generates empty summary if it couldnt find the keyword present in the input document which meets the attention weight beyond a threshold. Also, the proposed system is focused in maintaining the structure of the summaries generated for the given document.
  Keywords: text; summarize; document; recurrent.
  
  
• Web Server Workload Prediction using Time Series Model  
  by Mahendra Pratap Yadav, Akanksha Kunwar, Ram Chandra Bhushan, Dharmendra Kumar Yadav 
  Abstract: In distributed systems, multi-tier storage systems and cloud data-centers are used for resource sharing among several clients. To fulfill the clients request, the cloud providers share it's resources and manage the workload, which introduces many performance challenges and issues. One of the main challenges is resource provisioning in virtual machine (VMs or Container) since VMs are subjected to meet the demand of users with different profiles and Quality of Service (QoS). This proactive resource management approach requires an appropriate workload prediction strategy for real-time series data. The time series model exhibits prominent periodic patterns for the workload that evolves from one point of time to another with some short of time in random fluctuation. In this paper, a solution for the prediction of web server load problem has been proposed, which is based on seasonal ARIMA (Autoregressive Integrated Moving Average Model) model. ARIMA is a forecasting technique which predicts the future value based on its inertia. In seasonal ARIMA, seasonal AR and MA are used to predict the value xt (CPU workload time series) with the help of data values and errors at time lags that are multiple to the span of seasonality. We have evaluated our proposed method using real-world web workload data.
  Keywords: Cloud Computing; Elasticity; Auto-scaling; Time Series; Machine Learning.
  DOI: 10.1504/IJAIP.2022.10034175