International Journal of Intelligent Engineering and Systems

11

Authors: Hetal Vipul Gandhi, Vahida Zakirhusen Attar

Aspect Based Sentiment Analysis (ABSA) aims to perceive the opinion of the majority of reviewers for each feature (aspect category) of the product. We target both Aspect Term Polarity (ATP) and Aspect Category Polarity (ACP) subtasks of ABSA. We retrieve different word vector-based representations of sentence and augment them with n-gram and sentiment features using novel algorithm. We employ supervised machine learning approaches and propose the ensemble of Logistic Regression, k-Nearest Neighbor, and Support Vector Machine models for these subtasks. We experiment with four major domains from Hindi ABSA Dataset and report increase in performance accuracy by 16% for ATP subtask and 2% to 13% across domains for ACP subtask over results from researchers who released this dataset. Secondly, we also propose a novel WFBOS method to compute the Summary Score [range: 100 to 100] of each category from a particular domain. We make use of frequency of positive, neutral, negative, and conflict sentiment, of each aspect category from among the reviews of a particular domain in Hindi ABSA Dataset to do this computation.

31

Authors: Barlian Henryranu Prasetio, Edita Rosana Widasari, Fitra Abdurrachman Bachtiar

The nonverbal communication processes a critical parcel. In some cases verbal communication is incapable since the speaker does not utilize non-verbal communication well at the same time. Non-verbal communication which falls in unconscious emotion is important in determining in function of cognition, language comprehension, and decision making. However, a little research studied in this area. Many years, researchers are amazed by the reliability of Mel-Frequency Cepstral Coefficients (MFCC) feature extraction technique in recognizing stressed speech. In this paper, we propose a simple feature extraction technique that effective but strong enough to recognize stressed speech. There are the speech energy and frequency. We attempted a basic approach to classify unbiased or stretch on female and male discourse. The highlight extraction is based on control and recurrence. This investigate utilized 10 female and 10 male discourse datasets. There are 5 classification strategies utilized. The classification models are Neural Arrange, k-Nearest Neighbour, Bolster Vector Machine, combination of NN-k-NN and combination of NN-SVM. Test comes about approved utilizing k-fold cross-validation strategy. The tests are assessed utilizing R-index to distinguish whether the highlights contributing to the push discourse acknowledgment. Based on exploratory comes about, the number of inputs impacts the esteem of R-index. In general, combining the Neural Organize and Back Vector Machine is the most excellent classification strategy by appearing stretch acknowledgment rate of 85% precision.

53

Authors: Nischaykumar Hegde, Linganagouda Kulkarni

Mobile sink-based sensor networks deployed in unattended environment are characterized by infrequent sink visits. In these applications, ensuring the survivability and confidentiality of data is important till it is collected by the sink. The data must also be prevented from collection by agents other than sink and leakage by neighbours. Data collection in unattended environment must also be energy efficient to prolong the life time of the network. This work proposes a novel Energy Aware Adaptive Compressive Sensing (EA-ACS) assisted secure data survivability scheme which is able to optimize energy consumption at cost of data accuracy level desired by applications. The solution introduces three novelties: adaptive compressive sensing for reducing the foot print of data and energy consumption during transmission, a novel bi party authentication scheme for secure data collection by sink and low energy overhead piggy bagging for data replication. Through performance in NS2, the proposed solution is found to provide at least 6% higher data survivability and 20% lower energy consumption compared to existing works.

76

Authors: Puppala Ramya, Tumati Jashwanth, Dupaguntla Venkata Sathvik

Technological advancements in block chain (BC)-based frameworks have empowered scientists to create innovative inventions such as e-casting ballots. The traditional agreement models utilized proof-of-work (PoW) in the Bitcoin that affected energy utilization and bargained the adaptability for the ballot framework. The existing works evaluates the trust basically only on the centralized party as it was not feasible because of the dynamic changes in the pervasive social networking (PSN) topology and their characteristics. The present research work proposes a block chain based trust evaluation model for the PSN based BC. The proposed hybrid proof of stake-trust (PST) BC is based on the proof-of trust (PoT) and also the proof of stake (PoS) overcomes the issues that are occurring in the e-vote casting. The trust evaluation is performed for public verification and becomes transparent for each node of PSN. The advantage of the proposed method is that a new block will be designed for trust evaluation during the block generation. The process of sharding erases the workload when the network works fast for the individual nodes provides the sum of their alternative parts. Therefore, the model utilizes the agreements for generating the safe process that guarantee the precision to vote it from the time of the election results. The present research work utilizes the proof of stake-trust based BC resulted in security improvement. The model improves the adaptability and execution of the BC based on the ballot framework provided a secured voting system for the government. The proposed PST-BC model showed better results in terms of latency as 15/s when compared with the existing models merkle hash tree -bloom filter that obtained 107.3/s and performance constraints based electron of 18/s.

97

Authors: Krishna Reddy Gajulapalli, Merlin Sheeba Gnanadhas

The hybrid innovation of Orthogonal Frequency Division Multiplexing (OFDM) through Multiple-Input Multiple-Output (MIMO) delivers a feasible substitute to increase the Quality of Service (QoS) to accomplish better data rate and spectral efficiency for the transmission network. An extraordinary Peak-to-Average Power Ratio (PAPR) and Bit Error Rate (BER) are the important parameters that can be considered in the analysis of the MIMO-OFDM network. Partition Transmit Sequences (PTSs) is one of the capable procedures and direct methods to attain a reasonable PAPR performance. However, it needs a serious restoration process to discover the vital features that produce a computational multi-layered design with sub-blocks. Here, a reduced computational complexity PTS scheme is proposed which is completely depends on the hybrid procedure named as Moth Flame Optimization with Improved Firefly Algorithm (MFO-IFFA). The simulation outcomes demonstrated the capability of the proposed MFO-IFFA approach decreases the PAPR reduction up to 3.6 dB. Similarly, the proposed MFO-IFFA overcomes the significant evolutionary procedures mentioned in the existing works such as Additive Signal Mixing (ASM), Adaptive Simplified Optimized Iterative Clipping and Filtering (ASOICF) and Hybrid Independent Component Analysis (HICA).

116

Authors: Bharathi Ramachandra, T. P. Surekha

Wireless Sensor Network (WSN) is developed as a promising technology for a wide range of applications due to its low installation cost. However, the nodes in the WSN are susceptible to different security threats, because these nodes are located in hostile or harsh environments. The important goal of this paper is to obtain secure data transmission while minimizing energy consumption. In this paper, an Improved Moth Flame Optimization (IMFO) is used for selecting the optimal Cluster Head (CH) from the network. Moreover, the secure routing path is generated using the IMFO where the fitness function considers residual energy, distance, trust and node degree. Therefore, the Secure Clustering and Routing using IMFO (SCRIMFO) provides the resistance against the Distributed Denial-of-Service (DDoS) attack whereas the energy consumption is minimized by identifying the shortest path. The proposed SCRIMFO method is analysed in terms of Average Energy Consumption (AEC), Packet Loss Ratio (PLR), Packet Delivery Ratio (PDR) and routing load. Here, the SCRIMFO method is compared with the Secure Routing Protocol based on Multi-objective Ant-colony-optimization (SRPMA), Ant Colony Conveyance Algorithm (ACCA) and Energy-aware Trust and Opportunity-based Routing (ETOR) approaches. The PDR of the SCRIMFO is 94.64 % for 2 DDoS attacks that is high when compared to the SRPMA, ACCA and ETOR.

142

Authors: Shaymah Akram Yasear, Hayder M. A. Ghanimi

This paper proposes a modified honey badger algorithm (MHBA) for solving the optimal power flow (OPF) problem. This problem is a highly non-linear, non-convex and complex optimization problem with several decision variables and constraints. The original honey badger algorithm (HBA) has the problem of trapping in local optima due to the loss of population diversity, especially in solving complex optimization problems. Therefore, the MHBA aims at sufficient improvement in finding the optimal solution and feasibility. Opposition-based learning strategy (OBL) is integrated with the MHBA to preserve the diversity of the population and enhance the convergence toward the optimal solution. The effectiveness of the MHBA algorithm is evaluated on five objective functions of the OPF problem namely, total generation fuel cost minimization, active power and reactive power transmission losses minimization, voltage deviation and voltage stability enhancement. The performance of the proposed algorithm is tested and validated on the IEEE 30-bus test system. The proposed MHBA is compared with the HBA and other nature-inspired optimization algorithms reported in the literature. The results indicate that the proposed MHBA algorithm has the superiority to jump out of the local optimal and better convergence in solving the OPF problem. This is due to the strategy used in the algorithm which helps in maintaining the population diversity and provides a proper balance between exploration and exploitation.

167

Authors: Ammar Abdul Ameer Rasheed, Ahmed Sabah Al-Araji, Mohammed Najm Abdullah

During the previous years, the number of researchers who discussed the field of path planning algorithms has increased. In this paper, a proposed hybrid technique for path planning based on the benefits of sampling-based algorithms and heuristic path planning algorithms was enhanced. Rapidly-exploring Random Tree star (RRT*) was used to generate the candidate nodes that represent a collision-free path. These candidate nodes were applied to the Particle Swarm Optimization (PSO) algorithm to generate the shortest and the smoothest path planning (the optimal path planning) for the mobile robot. The RRT*PSO algorithm was applied in two scenarios (static and dynamic environments) with a workspace of [500×500] cm using MATLAB 2021a. In the static environment, a workspace full of obstacles was chosen to find the shortest collision-free path. In addition, a reference path equation was found to calculate the reference linear and angular velocities of the mobile robot as well as the linear and angular velocities of the right and left wheels. In this work, the suggested hybrid RRT*PSO algorithm improves path length 34.27% compared to the A* algorithm and the fuzzy analytic hierarchy process (A*–FAHP hybrid) algorithm and 0.35% compared to the Self-adaptive evolutionary game-based particle swarm optimization (SAEGBPSO) algorithm. In the dynamic environment, an algorithm named (contour path down and contour path up) was suggested to avoid collisions with dynamic obstacles. By using this suggested algorithm, the reference linear and angular velocities of the mobile robot as well as the linear and angular velocities of the right and left wheels were calculated to obtain a smooth path followed and free-navigation with dynamic obstacles in an environment.

198

Authors: Najla Akram AL-Saati

In this work, the wild horse optimization (WHO) algorithm, known for its ease of use, efficiency, and fast convergence, is explored in solving the reliability redundancy allocation problem (RRAP) for series-parallel systems. This problem has as of late caught the attention of researchers in this area, especially in today’s rapidly growing field of artificial intelligence. The NP-hard RRAP problem deals with maximizing of reliability under certain constraints. This work uses WHO algorithm to maximize the overall system reliability by determining how many redundant components are to be used along with their reliabilities in each subsystem, such reliability is constrained by cost, volume, and weight. Testing is carried out to show the effectiveness of this algorithm using four known numerical examples, results are to be compared with simplified swarm algorithm (SSO), attraction-repulsion imperialist competitive algorithm (AR-ICA), hybrid salp swarm algorithm and teaching-learning based optimization (HSS-TLBO), particle swarm optimization (PSO), and gradient based optimization (GBO). Computational results show that WHO was able to find better feasible near-optimal solutions effectively and efficiently in terms of population size and number of iterations.

216

Authors: Sritha Sreedharan, Ranjith Nadarajan

Dimensionality reduction is the most prominent process in artificial intelligence and data science because of using a massive amount of high-dimensional information. In recent days, many dimensionality reduction algorithms focused on the harmonic mean-based linear discriminant analysis (HLDA) which was the enhanced version of classical LDA. In particular, two different variants such as HLDA and HLDA pairwise (HLDAp) have been applied to reduce the high-dimensional data by using the harmonic mean between-class distance. However, its computation time complexity was high during the initialization phase since it comprises the matrix Eigen decomposition/inverse. Hence this article proposes the Fast HLDA (FHLDA) and FHLDA pairwise (FHLDAp) algorithms for reducing the high-dimensional data during classification. In this algorithm, a joint diagonalization scheme is introduced instead of Eigen decomposition depending on Taylor expansion for lessening the number of iterations in the initialization step to produce the discriminant. It does not a choice to a sweeping task so that every element of the Eigenvector matrix at every iteration is calculated to minimize the computation time burden. As well, a first-order approximation of the inverse Eigenvector matrix and the complete matrix of Eigenvectors are updated at every iteration. On the contrary, the overlap among the samples of dissimilar classes tends to miscategorization. So, the optimal discriminant vector is discovered to solve this issue by extending the between-class scatter matrices. Finally, the experimental outcomes show that the FHLDA and FHLDAp algorithms achieve 10 % higher average accuracy than LDA, ALDA, WLDA, G2DLDA, HLDA, and HLDAp algorithms.

237

Authors: Abdul Azeez, Suraiya Tarannum

In wireless communication systems, Multiple-Input Multiple-Out (MIMO) and Orthogonal Frequency Division Multiplexing (OFDM) were widely used technology. Spectrum utilization efficiency is needed to improve in OFDM technology for effective performance. A Nested Particle Swarm Optimization and Grey Wolf Optimization (HPSO-GWO) is proposed with the Low-Density Parity-Check (LDPC) coding scheme for minimizing the errors that occurred while transmitting the data bits over the MIMO-OFDM. The Nested PSO-GWO process involves in update the velocity of PSO using GWO method that helps to increases the convergence rate and balance the trade-off between exploration-exploitation process. The optimal velocity update in the PSO helps to escape from local optima trap in the search process and improve the efficiency of the model. Generally, the linear error-correcting codes are applied in Low-Density Parity-Check (LDPC) that is used to transmit the message in a noisy channel. Fast Fourier Transform (FFT), CP removal, and the Serial to Parallel (S/P) are accomplished as well as the received data bits are demodulated to obtain the output bits on the receiver side. The proposed method has 10-2 CCDF in 5.6 PAPR and the existing method requires more than 6 PAPR to achieve 10-2 CCDF.

257

Authors: Ahmed H. Aliwy, Ayad R. Abbas, Mustafa J. Hadi

Sentiment analysis (SA) is one of the most important tasks in the natural language processing (NLP) field. Many researchers have been trying to build an efficient SA system for many applications such as detection of terrorist activities, customer support management, analyzing customer feedback, market research, competitive research, and many others. Almost all these researches deal with a classification task, and they tried to solve one or two of the challenges that face this task. In this research, a complete SA system was constructed for processing five main challenges in SA and studying the effect of each one on the system. These challenges are the processing of negation, multi-polarity of words, text multi-polarity, semantically ambiguous words (exact meaning), and sarcasm. Some solutions were introduced for these challenges for getting high accuracy. It is the first work for studying the effect of these five challenges collectively with novel solutions to some of them. For aspect-based SA, three different classifiers were used; support vector machine (SVM), maximum entropy (MaxEnt), and long short-term memory (LSTM). The best results for f-measure were 0.833, 0.852, and 0.875 using the LSTM method on Foursquare ABSA, SemEval2014 (laptop and restaurant) datasets, respectively. We induced the most effective negation processing on the SA system from many test scenarios, followed by multi-polarity of words, semantically ambiguous words, text multi-polarity, and sarcasm, respectively. Also, the proposed sarcasm processing technique was evaluated on two annotated corpora with sarcasm, iSarcasm, and Rillof datasets, before applying it on SA.

279

Authors: Chada Sampath Reddy, G. Narsimha

Wireless sensor networking (WSN) is an important subject among the researchers in recent years and it is recognized as a familiar approach for upcoming applications like traffic monitoring in real time, battlefield, and eco-systems for surveillance. The network has receptive data and it is essential that the networks secure the several types of attacks like eavesdropping, capturing the node, rejection of service. Various sensor networks routing approach has been proposed to secure the networks. This work offers a secure routing scheme in WSN with deep learning approaches. Initially, trust verification is performed to select the non-malicious node, which is carried out through long short-term memory (LSTM), where the malicious nodes are removed. Secondly, data transmission is done through encrypting the data using elliptic curve cryptography (ECC) algorithm. This encrypted data is forwarded to the secured shortest paths, in which the secure shortest path is selected by dragonfly algorithm (DA) through solving the multi-objective function by incorporating the measures like distance, energy, delay, throughput, and trust. Evaluation of cost function of the proposed DA-LSTM model is correspondingly secured 22 %, 27 %, 31 %, and 18 % enriched performance than the PSO-LSTM, GWO-LSTM, WOA-LSTM, and CSA-LSTM at iteration 10th. Finally, the numerical simulations show that the suggested secure routing and data transmission in WSN is validated with other optimization algorithms and shows an advanced performance based on analysis with various performance measures by concerning with the number of sensor nodes and the maximum number of iterations.

303

Authors: Jatinder Pal Singh, Anuj Kumar Gupta

Wireless sensor networks (WSNs) have become a dominant technology in recent years due to their vast range of applications. However, the creation of energy-efficient WSN remains elusive. Clustering methods have long been employed in WSNs to reduce energy consumption and extend network lifetime, but efficient cluster head (CH) selection and competent routing always pose a major obstacle. This paper, proposes a new routing technique in which CH selection is performed using whale based tunicate swarm algorithm (WTSA) and routing between CHs is implemented using aquila optimizer (AO). In the proposed technique WTSA is used to identify efficient CHs from the WSN network, and remaining nodes join these CHs by assessing the cost function value. CHs residual energy, node degree, distance from base station (BS) and intra-cluster distance are being used to derive this cost function. AO is used for inter-clustering routing between CHs that are far from the BS. With the help of a fitness function, AO is used to cultivate optimal routes between CHs and to weed out ineffective routes. The performance of the proposed technique WTSA-AO is compared with state-of-the-art techniques BOA, PSO-GSA, BERA, WTSA, and LEACH in terms of delay, number of alive nodes, throughput and average energy consumption. The results show that the proposed technique improves the performance of wireless sensor networks by reducing average energy consumption by 8.3 % compared to WTSA and striving to achieve a delay of 0.121 seconds for 200 nodes.

329

Authors: Srinivasarao Thumati, S Vadivel, M Venu Gopala Rao

Several studies in the literature have focused the use of optimum network reconfiguration (ONR) and/or optimal allocation of renewable energy (RE) based distributed generation (DG) methodologies to improve performance in electrical distribution networks (EDNs). However, a few studies have only focused at simultaneous ONR and DG allocation in the context of demand increase, network loading unpredictability, or RE power. Furthermore, there is no much focus on the influence of emerging electric vehicle (EV) load penetration on EDN performance. This study proposes a novel meta-heuristic strategy for solving the simultaneous optimal ONR and RE-based DGs allocation problem for minimising distribution losses, improving voltage profile, and lowering greenhouse gas (GHG) emissions while considering the load penetration of electric vehicles (EVs). For solving the complexity inherent in the multi-objective optimization problem with numerous constraints and multi-type variables, a new and efficient honey badger algorithm (HBA) is proposed. On the IEEE 33-bus EDN, simulations are run with various EV load penetration levels. In terms of fitness function, HBA is observed to be superior to COA, TLBO, CSA, and PSO. As observed in the results, for 50% EV load penetration, the losses, AVD, and GHG emissions are increased to 1571.676 kW, 0.02658 p.u., and 25780.05 103 (lb/h), respectively. However, by having DGs optimally and with ONR, the losses, AVD, and GHG emissions are decreased to 77.08 kW, 0.00038 p.u., and 2175.01103 (lb/h), respectively. This indicates that the performance of EDN is improved significantly by having simultaneous DGs allocation and ONR solution irrespective of EV load penetration level. This also indicates the suitability of the proposed methodology for practical applications.

352

Authors: Harmandeep Singh, Vipul Sharma, Damanpreet Singh

Recently, breast cancer has achieved top position in terms of cause of mortality among women of all age groups by surpassing the lung cancer. To improve the survival rate, timely assessments are essential. Mammography is the best modality among the most widely used timely detection modalities. The radiologists’ manual reading may have an impact on the accuracy of the diagnosis. As a result, the computer-aided diagnosis (CAD) systems are being developed as tools to reduce the false alarms and to increase the diagnosis accuracy. In this study, an attempt has been made to improve the diagnosis performance of the CAD systems by incorporating recently developed marine predators algorithm (MPA) in conjunction with three different neural network classifiers including feedforward neural network (FFNN), cascade forward neural network (CFNN), and recurrent neural network (RNN). Unlike other existing studies, fully digital mammogram images from INbreast dataset have been employed for testing of the system proposed in this study. Experimental results reveal that the best classification performance [Accuracy 97.34%, Sensitivity: 98.40%, Specificity: 100.00%] is obtained when MPA is used in conjunction with RNN classifier. To demonstrate the usefulness of the proposed system, the obtained results are compared with the results obtained using already invented CAD systems in previously published studies using the same dataset. The findings suggest that the proposed system is acceptable for real-time clinical applications.

382

Authors: Suwito Suwito, Mochamad Ashari, Muhammad Rivai, M. Anis Mustaghfirin

This paper presents a multisector drip irrigation system (DIS) powered by solar photovoltaic (PV). A binary particle swarm optimization (BPSO) method possibly determines irrigation in multisector, depending on the power availability of the solar PV. The power from solar PV in this study was optimized through maximum power point tracking (MPPT) using an incremental conductance (INC) method. A prototype was built using a PV emulator in 10 controlled irrigation sectors. Three operating modes applied include control of all sectors with BPSO, partly sectors with a combination of linear programming, and all sectors with bypassed without BPSO. Simulations and tests showed very similar results, indicating that the BPSO method provides an optimal and accurate irrigation system with 0.7 psi of an average irrigation operating pressure (IOP) error, 4.01 % of water volume error, and 96.55 % of the uniformity coefficient of water volume.

406

Authors: Sudha Rani Chikkalwar, Yugandhar Garapati

An intrusion detection system monitors the networks and identifies the malware or suspicious activity in the network. Machine learning techniques were applied in the Intrusion detection system to improve its efficiency in the identifications. Imbalance data problem in intrusion detection affects the performance of identification and deep learning methods have overfitting problems. The autoencoder – support vector machine – grasshopper optimization (AE-SVM-GO) model is proposed to overcome the limitation of the overfitting problem in intrusion detection. The hybrid technique of AE-SVM-GO is applied to solve imbalance data problem and overfitting problem in intrusion detection. The autoencoder model is applied to generate the instances of minority classes to balance the dataset. The Grasshopper optimization performance hyper-parameter optimization in the SVM model to learn the features to adaptively set the parameter in classification. Four datasets such as UNSW-NB15, CICIDS2017, NSL-KDD, and Kyoto 2006+ dataset were used to test the proposed AE-SVM-GO model. The proposed AE-SVM-GO model has 95.3 % accuracy, whereas the existing convolutional recurrent neural network (CRNN) and SVM-naïve bayes model has 76.82 %, and 93.75 % accuracy respectively.

425

Authors: Ravi Mohan Sharma, Chaitanya P Agrawal

With the fast advancement of smartphone technology, the smartphone has emerged as the most prevailing instrument for accessing the Internet and obtaining a wide range of services with a click. Increased use of smartphones for online payments has attracted fraudsters, adding to an increase in malware outbreaks. Mobile application vulnerabilities and malware are the origins of various types of fraud and numerous cyber-attacks. Large datasets are frequently used for malware analysis; however, large datasets may contain many redundant, inappropriate, and noisy features, resulting in misclassification and low detection rates. This paper presents a hybrid approach to Android malware detection that reduces the dimensionality of datasets to reduce resource-intensive computation while preserving critical information. We present a novel hybrid approach for detecting Android malware based on a meta-heuristic (modified Intelligent Water Drop Algorithm (IWD)) and Deep Learning (DL) techniques. The studies show that the proposed approach efficiently removes irrelevant attributes and attains significant detection performance with an F1-Score of 93.7%, a precision of 95.35, an accuracy of 99.12%, and a recall rate of 96.68%.

447

Authors: Kavitha Muruganantham, Subbaiah Shanmugasundaram

Online digital marketing achieves their revenue according to their advertisements or sales assignment when companies have the profitable attention for recommending their products to customers via ranking them. Online customers are not able to guarantee that the items delivered through the recommendation by big data are either comprehensive or applicable to their essentials. In the past few years, recommendation frameworks were broadly applied to analyze the massive amount of data. Among those, a Distributed Predictive model with Matrix factorization and random Forest (DPMF) has achieved high efficiency to predict the item ratings. However, it accounts only for user preferences and opinions whereas the other contextual data are necessary for enhancing the efficiency of rating prediction. In this article, a Distributed Improved Predictive model with a Matrix factorization and random Forest (DIPMF) framework is proposed that considers the elements of social context and the dynamic characteristic of every user for every item to enhance the quality of prediction. The primary aim is to combine the information from the preferences, opinions and social context of each user. The social context of users is multiple features of the context such as differences in the current opinion with earlier opinion, behavior, relationship and interaction. Since each user is connected through relations and interactions. At first, the training dataset is split into an optimal amount of splits for accelerating the parallel and distributed training process. Then, the training process is carried out by the DPMF with the Distributed Improved Predictive model with Matrix factorization-Improved variant (DIPMI) to create the representation of every user's preferences, opinions and social contexts in the training set. Further, the prediction of rating is formulated as a regression challenge and solved via the Random Forest (RF) algorithm that predicts the customer’s rating behavior with their opinions and social context for every product. Finally, the experiments are conducted on trip advisor and Amazon datasets to evaluate the efficiency of DIPMI and DIPMF compared to the state-of-the-art recommendation frameworks. The findings exhibit that the DIPMF on the trip advisor dataset achieves an average of 0.6826 Root Mean Square Error (RMSE), 0.5925 Mean Absolute Error (MAE), 0.8369 recommendation quality and 0.0023 Confidence Range (CR) 95% compared to the other frameworks. Similarly, the DIPMF on Amazon dataset achieves an average of 0.7591 RMSE, 0.5704 MAE, 0.8298 recommendation quality and 0.0032 CR 95% compared to the other frameworks.

468

Authors: Binuja Philomina Marydasan, Ranjith Nadarajan

In mobile adhoc networks (MANETs), the most essential problem is to ensure the quality-of-service (QoS) during data transmission via multiple paths from a source to the destination node. To tackle this problem, a topological change adaptive adhoc on-demand multipath distance vector (TA-AOMDV) routing protocol was developed, which ensures QoS according to the different parameters for high-speed node mobility. But, it did not consider the path stability related to the node density, which also affects the routing efficiency. Hence, this paper develops a reliable and stable TA-AOMDV (RSTA-AOMDV) routing protocol that applies a new forwarding strategy to enhance the path reliability and stability for data transmission. In this protocol, every origin node relays data packets to the target in a hop-by-hop manner depending on local information gathered from its one-hop adjacent during the forwarding phase. Also, two different metrics are considered: (i) destination region selection (DRS) and (ii) weighted closeness and connectivity (WCC) to find the stable path from the source to the destination node. During the adjacent finding phase, a forwarding node self-selection mechanism is introduced to reduce the communication overhead due to the high node density and severe congestion. Finally, this protocol is simulated for 2 different cases: (1) a varying node mobility and (2) varying number of nodes. In the case of high-speed node mobility, i.e. 50m/s, the RSTA-AOMDV protocol achieves a 50.7 % packet delivery ratio (PDR), 109.4ms end-to-end delay (E2E-D), 158.51Kbps throughput, 73.4 % normalized routing overhead (NRO) and 11J mean energy consumption (MEC) compared to other existing protocols. Similarly, in the case of high node densities, i.e. when the number of nodes is 100, the RSTA-AOMDV protocol achieves a 66.5 % PDR, 1485.74ms mean E2E-D, 639.7Kbps throughput, 15 % NRO and 22.7J MEC compared to the existing protocols.

489

Authors: Valarmathi Srinivasan, Vijayabhanu Rajagopal

Diabetic Retinopathy (DR) is a common complication of Diabetes Mellitus (DM) that produces retinal abnormalities and can lead to blindness if not diagnosed and treated on time. To address this concern, an adaptive Convolutional Neural Network (CNN) model with Gradient Boosting (GB) called ResNetGB has been used from the literature where a Principal Component Analysis (PCA) based Fully Connected (FC) layer is used to capture the discriminative characteristics from the Retinal Fundus (RF) samples. It is essential to extract more effective features to categorize the DR grades. Hence, in this article, the Multi-Scale Attention (MSA) strategy is incorporated into the ResNetGB model for effective DR grade classification. First, the encoder network is used to embed the RF image in a high-level interpretational space in which the mixture of mid and high-level characteristics is considered to enhance the representation. Then, a Multi-Scale Feature Pyramid (MSFP) is added to define the retinal pattern in various localities and the MSA strategy is applied to the high-level interpretation. Moreover, the entire MSA-ResNetGB framework is trained by the cross-entropy loss to categorize the patients with respective DR grades. Finally, the experimental analysis exhibit that the MSA-ResNetGB model achieves the 94.40% and 94.17% accuracy on two benchmark datasets: Kaggle-APTOS and IDRiD, respectively compared to the cutting-edge models.

509

Authors: Muhammad Kaleem, Ismail Fauzi Bin Isnin

Due to technology scaling, network-on-chip (NoC) become the viable solution for on-chip many-core systems. The most critical concern of NoC is congestion management caused due to heavy communication traffic between nodes. Without an appropriate congestion resolution strategy for reducing heavy in-network traffic, the efficiency of the entire network is damaged severely. In this paper, an interval based record-keeping mechanism is presented to reduce network traffic and congestion by maintaining a history table and previous packet transaction records at each node. Proposed method performs certain validity checks before allowing using previous transaction record from history table. The performance of the technique is investigated in terms of average delay and compared to the state-of-the-art routing algorithms using the Access Noxim simulator. The simulation results demonstrate that the proposed method has outperformed in terms of global average delay, with 8-12% improvement, the average number of hits is 26-61% greater than misses under different synthetic traffic. The proposed algorithm has been tested under various topological configurations for efficiency evaluations.

533

Authors: Muhammad Aidiel Rachman Putra, Tohari Ahmad, Dandy Pramana Hostiadi

Botnets are a severe threat to a computer network, affecting various aspects of security systems, including spreading malicious programs, phishing, sending spam messages, and click fraud. Because of their negative consequences, botnets must be identified early. Nevertheless, their different characteristics have made them challenging to detect. This research proposes a bot patterns communication detection from traffic flows analysis consisting of three main activities: bot detection, extraction, and communication behavior analysis phases. This proposed model aims to obtain a specific behavior of bot attacks, which can be used as an early warning bots attack system. The process of bot patterns communication detection depends on the accuracy of bot detection, so the model improves the pre-processing phase and uses multi-model classification. Improvement in the pre-processing phase is carried out in the feature engineering section using the concept of one-hot encoding. Several machine learning classification models are used to obtain the best detection accuracy: Decision tree, Random forest, Logistic regression, k-NN, and Naïve Bayes. Furthermore, the model has been tested on two different datasets, namely the NCC and CTU-13. The experimental results show that the proposed model is optimal and recognizes bot activities well. The accuracy detection is obtained at 99.99%. Besides, the model can also identify the bot’s attack activity scenario and communication behavior in three types: centralized, distributed, and spread.

555

Authors: Salma El bakkal1* Abdellah Lakhouili1 El Hassan Essoufi

Intelligent traffic controller leads to manage traffic at intersection in order to minimize traffic congestion and has been intensively researched for a several decades. Multi-intersection cooperative traffic signal control (CTSC) is an efficient system that has received a great deal of attention and development in recent years. One problem with multi-intersection CTSC is that controller’s actions are based only on the traffic state or on the decisions taken at previous time step (t-1) at adjacent intersection. To address this problem, in this work a Double Deep Q Network Cooperative Traffic Signal Controller (CTSC-DDQN) is proposed. The CTSC-DDQN algorithm is a reinforcement learning agent that, depending on current traffic conditions, changes the traffic phase distribution order for n-intersection simultaneously. Experimental results under real scenarios show that the proposed approach outperforms other static approach which fixe the time length and the phase order despite the traffic situation and actuated controllers which change the traffic light properties based on the queue length in term of average Q-length and average waiting which eventually leads to mitigate traffic congestion. The results show that our could minimize the average waiting time by up to 79% and the average queue length by up to 80%.

575

Authors: Sai Ram Inkollu, G.V. Prasanna Anjaneyulu, Kotaiah NC, Nagaraja Kumari.CH

Currently global warming is increasing significantly around the world and become one of the typical issues for sustainability. In this connection, many sectors are transforming towards sustainable solutions. Integration of renewable energy (RE) is one such adaption in electrical power systems, by which the burden on conventional plants as well as their greenhouse gas (GHG) emission can be reduced significantly. However, due to radial structure, electrical distribution networks (EDNs) may not support for RE integration inappropriately. There are numerous methods have been introduced on optimal allocation of RE sources in radial distribution networks (RDNs), but not highly focused on maximizing their RE hosting capacity (HC). In this paper, a recent and efficient meta-heuristic algorithm, hunter-prey optimization (HPO) is introduced for finding the optimal locations and sizes of photovoltaic (PV) systems in RDNs. In addition to the loss minimization and voltage profile improvement, maximization of PV hosting capacity (HC) is focused as a major objective. Simulation results are done on IEEE 33-bus RDNs for different scenarios. The computational efficiency of proposed HPO is compared with other recent algorithms and it is observed that the results of HPO are better than other compared methods in terms of global optima. In addition, the enhanced HC of PV systems in RDNs is ensured improved performance in terms of reduced grid-dependency, GHG emission and distribution losses along with improved voltage profile.

599

Authors: Shatha Abdulhadi Muthana, Ku Ruhana Ku-Mahamud

This study presents a comparison between multi-objective optimization methods used to obtain solution for generator maintenance scheduling (GMS) problem. The GMS problem with three objectives which include the total operation cost minimization, system’s reliability (gross reserve) maximization and convenience is considered in this study. Convenience objective is represented by the minimization of the violation in the number of units’ maintenance outage constraint. To solve the problem of GMS, there is a need for GMS model to represent the requirements of electrical power system and optimization method to implement the model and obtain solution for the GMS. The proposed Pareto ant colony system (PACS) algorithm is based on ant colony system algorithm and Pareto approach. Pareto approach is used to make trade-off between the obtained solutions based on the three objectives. In this study comparison is made based on the results of experiments using the IEEE RTS 32- and 36-unit systems while the demand for the 32- and 36-unit systems is based on IEEE RTS demand systems. In addition, four common multi-objective algorithms based on Pareto approach i.e., the Non-dominated Sorting Genetic Algorithm II, Strength Pareto Evolutionary Algorithm 2, Multi-objective Simulated Annealing, and Multi-objective Particle Swarm Optimization are used in the evaluation of the proposed PACS algorithm. The multi-objective GMS model is implemented by all the algorithms and five performance metrics i.e., the grey relational grade (GRG), coverage, distance to Pareto front, overall Pareto spread and the number of non-dominated solutions are used in the evaluation. The Friedman test is also used to evaluate the algorithms’ performance statically, which is made based on GRG metric. The experimental results showed that the proposed PACS algorithm was able to obtain a robust solution by considering different initial operational hours of the units. In term of GRG metric, the PACS algorithm was able to obtain the best results for the 32-unit systems in all the maintenance windows. However, for the 36-unit system, the PACS algorithm secured the second-best results at the early stages of the operation time but outperformed other algorithms during other operation times. For other metrics, overall the PACS algorithm has the best performance in terms of coverage, distance to Pareto front and overall Pareto spread metrics while the NSGAII has the best result in terms of the number of obtained non-dominated solutions. Friedman test implies that the PACS algorithm is significantly better than the other comparative algorithms.

627

Authors: Saka Uma Maheswara Rao, K Venkata Rao, Prasad Reddy PVGD

The medical and health service has become advanced and the smart health care platform has made the diagnosis more robust for the treatment. The accurate analysis of medical data is dependent on early disease detection and the value of accuracy is reduced when the medical data quality is poor. However, the existing approaches failed to deploy the learning model to handle the heterogeneous medical data. The present research work has used the machine learning algorithm effectively for the chronic disease prediction such as heart disease, cancer, diabetes, stroke, and arthritis for the frequent communities and the medical data are available from the disparate sources including a wide variety of information has made difficult to analyse and retrieve. The detailed information about the attributes is required to be known as it is significant in analyzing the medical data. The process of selecting the attributes plays an important role in decision-making for medical disease analysis. The proposed model was tested for its effectiveness and was validated with various benchmark data which was collected from distinct domains. The present research work utilizes Spark Streaming layers for data streaming to diagnose further based on Long Short Term Memory (LSTM) Co-learning with whale optimization approach is from the heterogeneous medical data. The results obtained by the proposed method analysed the disease abnormality better. The existing model obtained SVM-RBF of 81.30 %, k-fold cross-validation and hyperparameters tuning with Random Forest of 94.9%, CNN of 90%, and the proposed LSTM based Co-learning model with Whale optimization approach obtained 98.6 % which is better compared to the existing models.