International Journal of Intelligent Engineering Informatics (10 papers in press)
Human Face Gender Identification using Thepades Sorted N-ary Block Truncation Coding and Machine Learning Classifiers
by Sudeep Thepade, Deepa Abin, R.I.K. DAS, TANUJA SARODE
Abstract: Human face gender identification is increasingly gaining popularity because of exponential popularity and interest in ubiquitous and pervasive computing. The computing embedded in the environment is pervasive computing, where the environment can be made to feel the person is present there and according to the person whether male or female the environment may induce certain decisions with help of ubiquitous computing devices to make environment suited to the person. The challenge of detecting a face is male or female is very trivial due to similarity of features of faces. The paper presents use of Thepades Sorted N-ary Block Truncation Coding (TSBTC N-ary) for extraction of features for the faces and further deploys various machine learning classifiers to identify the face as male or female. Here TSnBTC is explored for six different combinations (from two-ary to seven-ary). Fourteen machine learning classifiers are explored with each of these feature extraction generation methods for face gender identification giving 96 variations of the proposed method, all these are tested using Faces94 dataset. Classification accuracy is used as performance measure. Overall Random Forest has given best performance and TSBTC-7ary outperformed other feature extraction variations.
Keywords: Gender Identification; Machine learning Classifier; Thepade’s Sorted BTC N-ary; TSBTC N-ary.
A Hybrid Framework for Ranking Reviewers Based on Interval Type-2 Fuzzy AHP and VIKOR
by Hossein Abbasimehr, Esmaeil Nourani, Mostafa Shabani
Abstract: Online reviews are crucial resources both for users and business enterprises. However, the quality of online reviews varies greatly. To address the problem of low-quality reviews, we focus on the problem of reviewer credibility and propose a new framework. The proposed framework incorporates three main parts including identification of source credibility factors, preprocessing, and ranking via interval type-2 fuzzy analytical hierarchy process (IT2FAHP) and VIKOR method.
A major distinction of the proposed framework is utilizing multiple factors obtained from different sources, which leads to considerably improved reviewers ranking. This is in contrast to the other approaches which consider limited factors for credibility computing. Furthermore, we propose a new method for defuzzification of interval type-2 fuzzy sets that yields more reasonable results when compared with the existing methods. The last but not least is that it can be employed by companies to filter customers reviews based on their credibility rank. The framework is evaluated using real data crawled from Epinions. The results indicated that the proposed framework effectively ranks the reviewers based on their credibility.
Keywords: Online Reviews; Reviewer credibility; Interval type-2 fuzzy set; AHP; VIKOR.
Estimation of centroid, ensembles, anomaly and association for the uniqueness of human footprint features
by Kapil Kumar Nagwanshi, Sipi Dubey
Abstract: Present study proposes an approach for the estimation of the distinctiveness of human footprints under the machine learning environment. In this system, a sum of 880 raw footprints have been segmented to get the twenty-one features for ensemble learning. All the features have been analyzed for computation of minimum, mean gray value, median, maximum, standard deviation, kurtosis, and skewness for footprint dataset. The G-means clustering offers centroid information of footprint features. A set of ten ensembles has analyzed for surrogate footprint attributes. Ten anomaly models were created for anomaly scores among these features. The association of features gives the uniqueness of the human footprints for personal identification through fuzzy rules for every set of ensembles. As a consequence, centroid, ensembles, anomaly, and affiliation proved the individuality of human footprints.
Keywords: Anomaly Score; Association; Centroid; Ensemble; Footprint; G-means Cluster; Machine Learning; Recognition; Rule-base; Segmentation.
Detecting Intrusive Transactions in Databases using Partially-ordered Sequential Rule Mining and Fractional-distance based Anomaly Detection
by Indu Singh, Mononito Goswami, Rishabh Mathur, Minkush Manuja
Abstract: Databases are popular means to store, retrieve and query massive
quantities of transactional data critical to the regular functioning of organizations.The reliance of organizations on this data, coupled with the proliferation of the internet, has made databases prone to data breaches by disgruntled employees and hackers. Illegitimate access to databases may compromise their integrity and confidentiality, resulting in legal and financial ramifications for organizations. To this end, we propose a Database Intrusion Detection System (DIDS) called Fractional-distance based Anomaly Detection with Partially-ordered Dependency Analysis (FADPDA) to identify malicious transactions issued to databases. To weed out such transactions, our DIDS combines data dependency analysis using
security-sensitive Partially-ordered Sequential Rules (POSRs) with fractional-distance based anomaly detection. Data dependency rules capture sequential patterns in database access while the anomaly detection module builds profiles of regular transactions based on their syntactic features. Unlike most prior work, FADPDA can seamlessly run on both RBAC administered and non-RBAC databases. Detailed experiments on two databases a TPC-C benchmark and a synthetic database, revealed that POSRs effectively and efficiently represent data dependencies. Furthermore, combining data dependency analysis and anomaly detection reduces our systems reliance on hyper-parameters such as support and confidence thresholds, and enhances its intrusion detection capabilities. Through our experiments, we also show that our approach FADPDA outperforms major existing DIDS in terms of precision and recall values.
Keywords: Database Intrusion Detection; Anomaly Detection; Attribute Sensitivity; Sequence Reactivity; Partially-Ordered Sequential Rule Mining; Fractional Distance Metrics.
ROUGH TOPOLOGIES ON CLASSICAL AND BASED COVERING ROUGH SETS WITH APPLICATIONS IN
MAKING DECISIONS ON CHRONIC THROMBOEMBOLIC PULMONARY HYPERTENSION
by Nof Alharbi, Hassen Aydi, Cenap Ozel, Selçuk Topal
Abstract: The main aim of this study is to show how to evaluate classifications of data of Chronic Thromboembolic Pulmonary Hypertension (CTEPH)symptoms by using classical rough, classical rough nano, based covering rough and based covering rough nano
topologies respectively. The attributes are examined how effective they are in detecting the disease correctly when the strong and weak, internal and external points defined in the topologies are evaluated together with the symptoms. In addition, the methods used by topologies in the detection of symptoms were explained by pseudo algorithms
Keywords: Rough topology; rough nano topology; rough topology based covering nano topology; key attributes; decision making; strong interior; weak interior; weak key patients.
Recognition of flowers using convolutional neural networks
by Abdulrahman Alkhonin, Abdulelah Almutairi, Abdulmajeed Alburaidi, Abdul Khader Jilani Saudagar
Abstract: Every human has curiosity about what's around them. Most of the people love the nature and visits different places like parks, flower shows etc, with family and children during free time. But due to lack of enough knowledge and information it is very difficult to decide which flowers are beneficial, non-poisonous and edible to mankind. To solve this problem, this work developed a mobile application which capture flower images and helps in recognizing the flowers and categorize them into different categories using Deep learning algorithms. This work uses a dataset which contains four different flowers (Sunflower, Dandelion, Rose, and Tulip) for training purpose and tested with a sample of flowers over the trained model. The percentage of overall accuracy achieved in recognition of flowers is approximately 83.13%.
Keywords: flower recognition; deep learning; keras; mobile application; tensorflow.
Investigating the Sustainability of a Food System by System Dynamics Approach
by Alireza Amiri, Yahia Zare Mehrjerdi, Ammar Jalalimanesh, Ahmad Sadegheih
Abstract: Food is one of the essential human needs, from birth to death. The existence of sustainable systems that satisfy such requirements is always necessary for the survival of the human race. This study investigates the sustainability of a food system using the system dynamics approach. The proposed model is applied to a food system in Iran. For this, the key factors affecting the production of wheat, the primary food resource in Iran, were analyzed. Based on the history of these factors and related research works, causal loop diagrams were depicted. Then, the problem was simulated by depicting the flow diagram. After the model validation, several scenarios were proposed for sustainable wheat production. Reducing the use of chemical fertilizers and increasing the use of organic fertilizers, while in the short term reduce the production, in the long term, by improving soil quality, bring about an increase in cropping sustainability and wheat production.
Keywords: Food system; simulation; system dynamics; system sustainability; wheat production.
Dynamic strategies for measuring the performance of inventory system in closed loop supply chain
by Babak Shirazi
Abstract: Closed loop supply chains are the complicated dynamical system, which focuses on all the factors involved in the comprehensive process and seeks to adopt strategies for achieving the chain established targets. The purpose of this study is to evaluate the food supply chain performance. A system dynamic approach is applied to examine the relationships and behavior within the closed -loop supply chain. The variables with a significant effect on system performance are identified. Then, the simulated model is analyzed by sensitivity to provide a tool for managers to generate policies and business decisions which can lead to enhancing the performance of the supply chain system. The results reveal lost sale-oriented to zero also inventory cost and transferring time in the lowest state by adding an intermediary unit between segments and Implementation of the recommended policy.
Keywords: Dynamic system; Inventory management; Performance; Supply chain; Closed-loop.
Special Issue on: SUSCOM-2020 Advances in Computational Intelligence for Machine Vision and Image Processing
Dung Beetle inspired local search in artificial bee colony algorithm for unconstrained and constrained numerical optimization
by Nirmala Sharma, Harish Sharma, Ajay Sharma, Jagdish Chand Bansal
Abstract: In recent times, swarm intelligence (SI) centered strategies are proving their efficacy in the arena of engineering optimization problems. The artificial bee colony (ABC) algorithm is one of the efficient SI centered technique. To intensify the exploitation concept of ABC, a local search strategy inspired by dung beetle orientation and foraging activities is developed and amalgamated with ABC. This developed local search strategy is termed as dung beetle local search (DBLS) strategy. The developed amended algorithm is titled as dung beetle inspired ABC (DBABC) algorithm. The developed DBABC is analyzed using 32 unconstrained benchmark optimization problems, 18 constrained benchmark optimization problems, and 3 engineering design problems. The obtained outcomes validate the competitiveness of the proposed approach.
Keywords: Local search; Swarm intelligence; Dung beetle; Nature inspired algorithms; Artificial bee colony.
Low-Level Features based 2D Face Recognition using Machine Learning
by Sahil Sharma, Vijay Kumar
Abstract: In modern times, when deep learning-based face recognition is highly in demand, this paper presents machine learning techniques using the low-level feature extraction. Deep learning has a drawback of getting things done in a black-box, however extraction of low-level features viz. Histogram of Oriented Gradients (HOG), Speeded Up Robust Features (SURF), and Local Binary Patterns (LBP) with a machine learning-based classification model presents higher simplicity in understanding the concepts for a beginner. This paper presents the experimental demonstrations using twenty-two variations of machine learning models. Two face datasets, namely, Bosphorus and UMBDB present publicly, are used for evaluating different classification models. Four experimentations are shown in the implementation section to demonstrate the effect of feature extraction, discretization, feature variation, and adding noise in the image under probe. The Subspace Discriminant Ensemble model yields the highest efficiency in classifying faces using HOG features. In the final section of experimentation, the visual verification results are presented. The effect of various noise attacks on the probe image is shown in the last experimentation.
Keywords: Feature Extraction; Discretization; Classification; Face Recognition.