International Journal of Intelligent Systems Design and Computing (11 papers in press)
An Improved Harmony Search based Functional Link Higher Order ANN for Non-linear Data Classification
by Bighnaraj Naik, Janmenjoy Nayak, Himansu Sekhar Behera, Ajith Abraham
Abstract: To obtain the optimal set of weights in any higher order artificial neural network, it is often laborious to adjust the set of weights by using appropriate learning algorithm. In this paper, an improved variant of Harmony Search (HS), called Improved Harmony Search (IHS) along with Gradient Descent Learning (GDL) is used with Functional Link Artificial Neural Network (FLANN) for the task of classification in data mining. The IHS performs better than HS by eliminating constant parameters (Bandwidth (bw), Pitch Adjustment Rate (PAR)) in HS algorithm and incorporating changes dynamically in PAR and bw with iteration number. The searching capability of IHS to obtain optimal harmony in a population of harmony memory is used along with Gradient Descent Learning (GDL) to discover optimal set of weights for FLANN model. The proposed IHS-GDL-FLANN is implemented in MATLAB and compared with other alternatives (FLANN, GA based FLANN, PSO based FLANN and HS based FLANN). The IHS-GDL-FLANN is tested on various 5-fold cross validated benchmark datasets from UCI Machine Learning repository. In order to get statistical correctness of results, the proposed method is analyzed by using Friedman Test, Holm and Hochberg Procedure and Post-Hoc ANOVA Statistical Analysis (Tukey Test & Dunnett Test) under the null-hypothesis. The performance of IHS-GDL-FLANN outperforms other counterparts and is found to be statistically significant.
Keywords: Improved Harmony Search; Gradient Descent Learning; Functional Link Artificial Neural Network; Data Mining; Classification; Machine Learning.
Periodic Pattern Mining in Weighted Dynamic Networks
by Anand Gupta, Hardeo Kumar Thakur, Anshul Garg
Abstract: Graph is one of the media to represent and summarize interactions in a time varying network. Often, interactions repeat after a fixed interval of time and exhibit temporal periodicity. Existing algorithms focus either on the structure or on the weight of periodic interactions individually. But, for instance, a stock analyst requires evidence of both structure and weight (here, price) of the stock pairs to make prediction and discovers information about the profit producing stocks and the actual profit. On performing experiments using existing algorithms explicitly, it is observed that the efficiency is lost in such applications. Hence in this paper, we provide an efficient framework based on available algorithms to mine periodic patterns both on structure and weight in a weighted dynamic network. The proposed framework consists of a mapping between interactions that are periodic on structure and weight. We have performed experiments on synthetic and real world datasets. The results validate the scalability and practical feasibility of the proposed framework.
Keywords: Dynamic graph; Periodic graph; frequent graph; weighted graph.
Optimal Control of Cyber Physical Vehicle Systems
by Kaustav Jyoti Borah
Abstract: - The Cyber-Physical System (CPS) is a term describing a broad range of complex, multi-disciplinary, physically-aware next generation engineered system that integrates embedded computing technologies (cyber part) into the physical world. Cyber-physical systems (CPS) are physical and engineered systems whose operations are monitored, coordinated, controlled and integrated by a computing and communication core. Just as the internet transformed how humans interact with one another, cyber-physical systems will transform how we interact with the physical world around us. A cyber-physical system (CPS) is composed of tightly-integrated computation, Communication and physical elements. Cyber-physical vehicle systems (CPVSs) are advancing due to progress in real-time applications, control and artificial intelligence.Multi-objective design optimization maximizes CPS efficiency, capability and safety, while online regulation enables the vehicle to be responsive to disturbances, modeling errors and uncertainties. CPVS optimization occurs at design-time and at run-time. We will survey the run-time cooperative optimization or co-optimization of cyber and physical systems, which have historically been considered separately. A run-time CPVS is also cooperatively regulated or co-regulated when cyber and physical resources are utilized in a manner that is responsive to both cyber and physical system requirements. This paper surveys research that considers both cyber and physical resources in co-optimization and co-regulation schemes with applications to mobile robotic and vehicle systems. Time-varying sampling patterns, sensor scheduling, anytime control, feedback scheduling, task and motion planning and Resource sharing are examined. Cyber-physical systems will transform how we interact with the physical world around us. Many grand challenges wait in the economically vital domains of transportation, healthcare, manufacturing, agriculture, energy, defense, aerospace and buildings. The design, construction and verification of cyber-physical systems pose a multitude of technical challenges that must be addressed by a cross-disciplinary community of researchers and educators.
Keywords: New frontiers; control; real-time control; optimization; optimal Control; robotics.
KMeansAtkinson Clustering approach for Collaborative Filtering based Recommendation System
by Surya Kant, Tripti Mahara
Abstract: The amount of information generated by web is growing rapidly every day and results in information overload. The information overload problem makes recommendation system necessary. Collaborative filtering is one of the most successful approaches to design a Recommendation system. The key idea of this technique is based on the common interest of users. If the user has similar taste in past for a set of items, then they will share common taste in future. However, there is some weakness like sparsity, scalability etc. of this prosperous approach affecting the quality of recommendations. In this paper, a clustering based recommendation algorithm is proposed. The clustering technique has been used to form neighborhoods (groups of users who have similar preferences) of active user. It exploits underlying data correlation structures to choose the initial centroid for k-means. Experimental results on three benchmark datasets, MovieLens 100k, MovieLens 1M and Jester, demonstrates that proposed method exhibit superior accuracy in comparison to the traditional k-means based recommender systems.
Keywords: Collaborative Filtering; Recommendation system; Information filtering; Clustering; k-means.
A Prediction-based Handoff scheme for QoS in WLAN Systems
by Sanjay Biswash, Pavan Mishra
Abstract: In this paper, we introduce a technique to reduce the authentication overhead for handoff process in wireless local area networks (WLAN). It is based on user mobility prediction to reduce the network scanning area during the handoff procedure. The central server facilitates an intelligent mobility prediction mechanism to identify the next possible mobile access points for mobile subscriber. The predicted information is manage by an event log table, and it help to reduce the authorization overheads for cellular networks. It follows the reserve channel scheme to maintain the QoS for users during the call and handoff procedure and the handover-call has higher priority over local initiated call. The analytical model shows the effectiveness and efficiency of proposed work, we apply the Markov random walk model and M/M/1 queuing model to formulate our work. It compared with the tradition technique, and has 40% better result than legacy techniques.
Keywords: Wireless access point; Handoff management; Pre-authentication; Mobility prediction; Network overheads.
Stochastic Modeling for Traffic Flow: A Review
by Kaustav Jyoti Borah
Abstract: This paper presents a brief introduction to the history of traffic flow theory and reviews a modeling approach to traffic flow, specifically during breakdown flow from a previous IEEE paper which discuss the process the application to predict travel time reliability. This paper discusses the history, definitions, fundamental relationships and type of models. The stochastic process and estimation from the review paper for the breakdown flow contains two random flow breakdowns variables. Breakdown flow happens when the two random variables, breakdown speed and breakdown duration, occurs at the homogeneous level. The model follows microscopic model which are homogeneous and continues in a heterogeneous level for a macroscopic model. The models are based from the vehicle changes at different levels of their speed. The mathematical process used in order to obtain the random variables are Weibull probability distribution for the breakdown speed and also the probability density function for the breakdown duration as an estimation also known as the hazard function. Simulation using Monte Carlo method will compute the random sampling from the breakdown variables and breakdown duration into a macroscopic model. The results will compare the actual data collected from Caltrans in 2007 and the mathematical model and simulation will be the same traffic flow behavior and characteristics.
Keywords: breakdown duration; breakdown speed; breakdown flow; hazard function; macroscopic model; microscopic model; Monte Carlo Simulation; Weidull probability distribution.
A survey of detecting pedestrians from low resolution imagery
by Kyaw Kyaw Htike, Siew Chin Neoh, Zaw Zaw Htike, Choo Wou Onn
Abstract: Being able to detect pedestrians well in image or video has numerous potential benefits in many diverse applications such as image retrieval, surveillance systems, elderly monitoring for safety, crowd analysis, person counting, activity and behavior detection, and advanced driver assistance systems. Although there have been much work and research done in the field of pedestrian detection, recent state-of-the-art research has made it clear that a lot of improvements need to be made, especially when it comes to low resolution imagery. Despite a number of review papers that have been published that survey pedestrian detection, there is a great need to have a survey paper that focuses on pedestrian detection specifically for low resolution data. In this paper, we identify and break down the pipeline for low resolution pedestrian detection systems and survey its different components, as well as, presenting, and analyzing the underlying causes behind low resolution data as well as the potential solutions. In addition, we perform an in-depth critical analysis and review of the most representative and relevant papers in this area, also highlighting the strengths and weaknesses of different methods where relevant. We also discuss and make connections between the surveyed papers, give recommendations and outline future research directions which should be beneficial to both practitioners and researchers in this field.
Keywords: low resolution; detecting pedestrians; object detection; image analytics; computer vision.
Adaptive Backstepping Control for a Class of MIMO Uncertain Underactuated Systems with Input Constraints
by Ajay Kulkarni
Abstract: This paper presents a backstepping methodology basedrnadaptive controller scheme for a class of multi-input multi-outputrn(MIMO) uncertain underactuated systems in presence of actuatorrnconstraints. To develop a feasible controller scheme for multi-inputrnmulti-output underactuated systems, (n − p + 1) dimensions of thernn dimensional configuration space are stabilized by using onerndimension of the input space. This control term is developed byrnapplying hierarchical methodology whereas as remaining p − 1 inputrndimensions are assigned as dedicated control terms to solve therncontrol problem of remaining dimensions of the configuration space.rnBackstepping technique is used to develop the classical control termrnwhereas wavelet networks are used to approximate the uncertainrndynamics as well as to compensate the nonlinear effects of actuatorrnsaturation. A robust control term is used to attenuate thernapproximation error to a prescribed level. Uniform ultimaternboundedness (UUB) stability of the closed loop system is verified inrnthe Lyapunov sense. Simulation results illustrate the effectiveness ofrntheoretical development.
Keywords: Underactuated systems; hierarchical controlrnstructure; backstepping control; wavelet neural network,rnactuator saturation.
KNOWLEDGE SYSTEM FOR EARLY PHASE AESTHETIC CONCEPT GENERATION IN INDUSTRIAL DESIGN
by Sitaram Soni, Pritee Khanna, Puneet Tandon
Abstract: The early phase of the aesthetic concept generation involves the tacit knowledge of the experts. There is general lack of formal models to capture and use this knowledge, as it is difficult to externalize, capture, express and reuse. This paper contributes to the development of formal models and an application framework for the knowledge involved in early phase aesthetic concept generation of industrial products. The models are based on four axioms. These axioms are used to develop two models; aesthetic design complex (ADC) and action grammar. These models are used to develop a design learning and generation framework. Soft computing techniques are used to capture, reuse and externalize the tacit aesthetic design knowledge. The tacit design is expressed as heuristics, which are validated by human based evaluation. The developed prototype framework shows that such a computational support is possible to aesthetic design process, practice and education.
Keywords: design for aesthetics; knowledge based system; cognitive process; action grammar.
Prediction and Estimation of Civil Construction Cost using Linear Regression and Neural Network
by Nagaraj Dharwadkar, Sphurti Arage
Abstract: Adequate construction cost estimation is a main factor in any type of construction projects. Forecasting cost of construction projects can be considered as a difficult task. In order to forecast the cost of the civil construction projects, we have used the Ordinary Least Square Regression (OLSR) model and Multi-Layer Perceptron (MLP) in our proposed model. The performance of the proposed model is analysed on the data of the 12 years of schedule rates of construction projects in Pune region of India. The experiment shows 91% to 97% of accuracy in prediction using Ordinary Least Square Regression model. Similarly, we have conducted series of experiments on Multi-layer Perceptron model with different activation functions. It was observed that the Multi-layer Perceptron model with softplus activation function can be able to predict the project cost of the civil constructions with accuracy of 91 % to 98%. Thus it shows that the prediction of cost using Multi-Layer Perceptron model gives higher accuracy than the Ordinary Least Square Regression model.
Keywords: Construction Cost estimation; Ordinary Least Square Regression (OLSR); Multi-layer Perceptron (MLP); Activation functions; Root Mean Square Error (RMSE); Mean Absolute Percentage Error (MAPE).
Some Statistical Aspects of Children with Disabilities in Assam, India
by Jumi Kalita
Abstract: Analysis of data plays a very important role in describing a data set from various angles of interest. It digs out different characteristics intrinsic in the data set. This paper analyses the occurrences of disability in children for genderwise distribution, rural-urban distribution, and probes any probable relationship of mothers age with occurrence of disability among newborns. The influence of the parameters like birth-cry, birth-weight, mothers health during pregnancy and severe health problem of the children within a short period after the birth in determining the disability types are analyzed through multinomial logistic regression. Analysis of data relating to children with disability may be useful in predicting disability and taking precautionary measures.
Keywords: Disability; chi-square test; multi-nomial logistic regression.