International Journal of Intelligent Engineering and Systems

12

Authors: Risanuri Hidayat, Anggun Winursito

Research on the current speech recognition system leads to the creation of a noise-resistant system. The Mel Frequency Cepstral Coefficients (MFCC) extraction method becomes a popular method in the speech recognition system. In this paper, the MFCC's weakness of noise interference is the main reason underlies the accomplishment of a robust speech recognition system. Development was carried out by improving the denoising performance using a wavelet transform. Modifications were carried out by analyzing the weakness of the wavelet denoising process on the recognition system using the MFCC method. The analysis was conducted at one of the MFCC stages, the Fast Fourier Transform (FFT) stage. The proposed method was conducted by performing the denoising process using Wavelet only on the noise-related data based on the FFT process' analysis results. The study utilized speech data in the form of eleven isolated words in English added with noise with several different characteristics. Results showed that the proposed method was capable of generating a better accuracy than conventional wavelet denoising methods on the signal to noise ratio (SNR) of 10dB, 15dB, and 20dB using a Fejer Korovkin 6 wavelet type. The highest accuracy increase of the proposed method was in signal to noise ratio (SNR) of 15dB with a rise of 4.63%, followed by a 3.96% increase at 20dB intensity, and 2.3% at 10dB intensity. The performance of the proposed method is then compared with other methods. The results show that the proposed method has the best performance on clean speech and noisy speech at SNR intensities of 10dB, 15dB, and 20dB.

32

Authors: Karunya Rathan, Susai Michael Emalda Roslin

Wireless Mesh Networks (WMNs) have been considered one of the main technologies for configuring wireless machines since they appeared. In a WMN, wireless routers provide multi-hop wireless connectivity between hosts on the network and allow access to the internet through the gateway routers. These wireless routers are normally equipped with the multiple radios in the wireless mesh network that operate on multiple channels with the multiple interference, which is caused to reduce the network performance and end-to-end delay. In this paper, we proposed an efficient optimization algorithm to solve the channel assignment problem which cause due to the multichannel multi-radios in WMN’s. The main objective of our paper is to minimize the channel interference among networked devices. So, initially we construct a multicast tree with minimum interference by using Q-Learning algorithm, which is helps to minimize the end-to-end delay of packet delivery. From the constructed multicast tree, we intend to develop a channel assignment strategy with the minimum interference by using Modified version of Alternative Direction Method of Multipliers (MADMM) optimization algorithm, which is helps to increase the network throughput and packet delivery ratio. The proposed strategy was implemented by using NS-2 (Network Simulator-2) and the experimental result show that the performance of the proposed method is very high compared to the other method and the performance was calculated by using the feature metrics such as average throughput, packet delivery ratio, end-to-end delay and total cost, which is compared with the other existing channel assignment strategies such as Learning Automata and Genetic Algorithm (LA-GA), GA-based approach, link-channel selection and rate-allocation (LCR) and learning automata based multicast routing (LAMR) channel assignment methods.

54

Authors: Kumar Reddy Govindaswamy, Shriram Ragunathan

Genre Classification of movies is useful in the movie recommendation system for video streaming applications like Amazon, Netflix, etc. The existing methods used either video or audio data as input that requires more computation resources to process the data for the genre classification of movies. In this study, the Hierarchical Attention Neural Network (HANN) is proposed for genre classification of movies based on the social media called Twitter data as input. Twitter data related to the Telugu and English movies are collected and applied to HANN for movie’s genre classification. IMDB data are used to evaluate the performance of the proposed HANN method. The hierarchical structures of the twitter data is considered by the proposed HANN method and the most important words related to genre classification is identified by the attention mechanism, where the other neural networks such as Artificial Neural Network and Convolutional Neural Network (CNN) returns only the important weights resulting from previous words. The HANN method has the advantages of encoding the relevant information that helps to improve the performance of the recommendation system. The experimental results show that the HANN method achieve higher performance compared to other classifiers Long Short-Term Memory (LSTM) and Bidirectional LSTM (Bi-LSTM). The HANN method achieves accuracy of 73.15% in classification, while the existing BiLSTM method achieve the accuracy of 68% in classification.

72

Authors: Sridhar Ramaiah, Guruprasad Nagraj

Cluster head selection enacts a prominent role in Wireless Sensor Network to optimize the energy usage during the data collection. Few research works have been designed to choose the best cluster head in wireless network using different optimization techniques. However, cluster head selection performance of conventional algorithms was lower to extend the lifetime of network. Therefore, a Deep Neural Glowworm Swarm Optimized Soft C-Means Clustering (DNGSOSCC) model is proposed. Initially, DNGSOSCC model obtains number of sensor nodes as input at the input layer. After taking input, soft clustering process is carried in DNGSOSCC model at the first hidden layer where it groups the sensor nodes in WSN into a different cluster. Then, the glow worm population is initialized in DNGSOSCC model with the support of number of clusters formed at the second hidden layer. Next, Luciferin value is assessed in DNGSOSCC model for all sensor nodes according to their residual energy level in the third hidden layer. Followed by, the sensor node with higher luciferin value within cluster is selected as optimal cluster head at the fourth hidden layer to perform energy efficient data gathering in WSN. Finally, the output layer provides the selection result of optimal cluster heads in WSN. By using the above process, DNGSOSCC model significantly gathers the data from its cluster member through selected optimal cluster heads with minimal amount of energy. From that, DNGSOSCC model upsurges the lifespan of network through performing efficient route discovery and data collection in WSN. The DNGSOSCC model conducts the simulation work using metrics such as data deliverance ratio, data loss ratio, data transmission delay and network lifetime with reverence to number of data packets and sensor nodes. From the experiments conducted, DNGSOSCC model improves the data deliverance ratio and network lifetime by 19% and 10% as well as minimizes the data loss ratio and data transmission delay by 76% and 34% respectively than existing methods.

93

Authors: Mallikarjuna Anantapur, Venkanagouda Chanabasavanagouda Patil

Mobile ad hoc network is a self-organizing wireless network, which is formed by the wireless mobile devices without any centralized infrastructure. The mobile nodes in the network are mainly affected by the inherent characteristics such as unpredictable network topology and open wireless medium. Especially, the presence of selfish nodes in the network creates the packet loss and affects an entire communication system. In this paper, the hash function with position updating algorithm is proposed in the Ad hoc On-Demand Distance Vector (AODV) routing protocol for improving the security against the selfish nodes. The AODV routing protocol is used to transmit the data packets from the source to the destination. Therefore, the Prevention of Selfish Node using Hash Function (PSNHF) with position update algorithm is proposed for minimizing the packet loss through the network. The performances of proposed AODV-PSNHF method are analysed in terms of energy consumption, throughput, Packet Delivery Ratio (PDR), packet loss and normalized routing load. In addition, the AODV-PSNHF method is compared with the existing trust-aware ad-hoc routing protocol (T2AR). The selfish node identification using hash function and positioning update using AODV-PSNHF method provides reliable and secure data transmission under selfish nodes and shows better performance in terms of throughput, packet delivery ratio and packet loss. For 2% of malicious nodes the PDR of the AODV-PSNHF method is 89%, it is 9% higher when compared to the T2AR protocol whose PDR% is only 80% for the same.

113

Authors: M. Ganesh Karthik, M. B. Mukesh Krishnan

Internet of Things (IoT) has become more familiar in all applications and industrial fields such as medical, military, transportation, etc. It has some limitations because of the attack model in the transmission or communication channel. Moreover, one of the deadliest attacks is known as a Distributed Denial of Service Attack (DDoS). The Presence of DDoS in network layer cause huge damage in data transmission channel that ends in data loss or collapse. To address this issue the current research focused on an innovative detection and mitigation of Mirai and DDoS attack in IoT environment. Initially, number of IoT devices is arranged with the help of a novel Hybrid Strawberry and African Buffalo Optimization (HSBABO). Consequently, the types of DDoS attacks are launched in the developed IoT network. Moreover, the presence of strawberry and African Buffalo fitness is utilized to detect and specify the attack types. Subsequently a novel MCELIECE encryption with Cloud Shield scheme is developed to prevent the low and high rate DDoS attack in the Internet of Things. Finally, the proposed model attained 94% of attack detection accuracy, 3% of false negative rate and 5.5% of false positive rate.

134

Authors: Adi Wijaya, Teguh Bharata Adji, Noor Akhmad Setiawan

Electroencephalogram (EEG) based motor imagery (MI) classification requires efficient feature extraction and consistent accuracy for reliable brain-computer interface (BCI) systems. Achieving consistent accuracy in EEG-MI classification is still big challenge according to the nature of EEG signal which is subject dependent. To address this problem, we propose a feature selection scheme based on Logistic Regression (LRFS) and two-stage detection (TSD) in channel instantiation approach. In TSD scheme, Linear Discriminant Analysis was utilized in first-stage detection; while Gradient Boosted Tree and k-Nearest Neighbor in second-stage detection. To evaluate the proposed method, two publicly available datasets, BCI competition III-Dataset IVa and BCI competition IV-Dataset 2a, were used. Experimental results show that the proposed method yielded excellent accuracy for both datasets with 95.21% and 94.83%, respectively. These results indicated that the proposed method has consistent accuracy and is promising for reliable BCI systems.

157

Authors: Hadeel N. Abdullah

The diverse engineering and scientific applications are stated through complex and high-order systems. The significant difficulties of these systems are the complications of modeling, analyzing, and controlling. It is easier to examine simpler models for more physical insights than more complex models and result in lower-ordering controllers that are easier to implement. The model order reduction (MOR) was used to simplify the computational difficulty of such complications and was later developed intensively for use with increasingly CDS. In this paper, a new Modified Chaos Particle Swarm Optimization (MCPSO) technique is employed to get a reduced-order model of a large scale system and design a Linear Quadratic Regulator (LQR) based controller. The mod uses the combination of advantages of basic PSO algorithms and chaotic algorithms. It becomes an excellent algorithm with fast convergence, few control parameters, simple execution, and avoidance of local extremes. In addition to combining the chaotic algorithm, CPSO also improved the weight parameter w, adjusting it to the dynamic attenuation direction. First, efficient reduced-order model parameters are obtained for original higher-order systems based on the MCPSO. Then linear quadratic regulator (LQR (controller parameters optimized for the reduced-order model. The goodness of the proposed method is evaluated through a numerical example. The experimental results indicate that the proposed technique’s reduced order model provides an excellent close approximation to the original system.

181

Authors: Kasim K. Abdalla, Sameer Abdul Kadhim Alrufaiaat

A new robust decoding technique which designed of Multiple-Input Multiple-Output Space–Time Block Code (MIMO-STBC) using Fast Independent Component Analysis (Fast-ICA) based on proposed mixing model has been performed in this paper. This decoding technique is characterized by i) complexity is very low, ii) the speed is high and iii) BER performance is excellent. It can be achieved with any MIMO STBC system with a fewer pilot symbols number. Also, it is reduced decoding time into 1/8 by innovating a simple strategy referred by one source extraction method. Also, this paper includes suitable initializing for the de-mixing vector to solve the ambiguities problem of sign and source of blind source separation (BSS). To test the proposed technique, four transmitters (4Tx) STBC MIMO system was implemented using MATLAB2018. It also found that excellent BER performance associated with a high number of symbols per frame (about 8012 symbols). The simulation results show that the new decoder works for any number of receiver antenna (Nr = 2, 4 and 5). As compare with classical decoding algorithm, it is found that the new decoder provides coding gain (at BER =10-6) equal to 1 dB,1.45 dB and 1.76 dB when Nr = 2,4 and 8 respectively, using only 2~3 iterations only.

203

Authors: Nouf Saeed Alotaibi

Noise may affect images in many ways during different processes. Such as during obtaining, distribution, processing, or compressing. The Sparse Representation (SR) algorithm is one of the best strategies for noise reduction. One meta-heuristic algorithm is the Particle Swarm Optimization (PSO). This research demonstrates excellent results in noise reduction in the Fast PSO version while utilizing the SRs as well as meta-heuristic algorithms to gain. This method is known as FPSO-MP and it depends on the Pursuit Algorithm (MP) that matches. In this proposed study, a Dynamic-Multi-Swarm (DMS) method and a pre-learned dictionary (FPSO-MP) approach is presented to reduce the time for the learning dictionary calculations. The output of the denoising algorithm QPSO-MP is dependable on dictionary learning because of the dictionary size or increased number of patches. Similar to this work, a Non-locally Estimated Sparse Coefficient (NESC) is one explanation for the low efficiency of the original algorithm. Compared to the original PSO-MP method, these enhancements have achieved substantial gains in computational efficiency of approximately 92% without sacrosanct image quality. After modification, the proposed FPSO-MP technique is in contrast with the original PSO-MP method. The scientific findings demonstrate that the FPSO-MP algorithm is much more efficient and faster than the original algorithm, without affecting image quality. The proposed method follows the original technique and therefore reduces during run-time. The result of this study demonstrates that the best-denoised images can always be accessed from the pre-learned dictionary rather than the learning dictionary developed across the noisy image during runtime. We constructed images dataset from the BSD500 collection and performed a statistical test on these images. The actual findings reveal that the suggested method is excellent for noise reduction (noise elimination) as well as highly efficient during runtime. The analytical findings indicate that both quantitative and image performance outcomes are obtained with the proposed FPSO-MP approach during its contradiction with when denoising algorithms.

221

Authors: Andi Setiawan, Hajime Miyauchi, Ardyono Priyadi, Maurdhi Hery Purnomo

This paper presents the research results on a small scale compressed air energy storage (SS-CAES) operated at its maximum efficiency value in the discharged phase. This study used two methods to find the maximum efficiency; Standard Perturb and Observe (Standard P&O), and Multiple Input Multiple Output Fuzzy Perturb and Observe (MIMO Fuzzy P&O). The essential ideas of this study are to operate SS-CAES in its maximum efficiency in specific loading power and to improve the performance of standard P&O method in achieving maximum efficiency and reduce the oscillation amplitude of the efficiency value as well. To do this, in this study, we use two devices, namely a motorized valve to ensure sufficient power required for the load and a boost converter whose duty cycle will be adjusted so that the system can operate at its maximum efficiency. The study was conducted using a simulation model and compared these two methods performance for the same loading case. The experimental results reveal that the addition of the MIMO Fuzzy algorithm can improve the performance of the Standard P&O method in accelerating maximum efficiency achievement and reducing oscillations when the system has reached the maximum operating points. The performance improvement on the maximum efficiency point tracking (MEPT) in SS-CAES operations, evidenced by the increase in the average efficiency value in the given loading scenario of about 1.22%. Besides, the application of MIMO Fuzzy P&O also impacts reducing the amplitude of speed oscillations in the steady-state area, which is about half of the amplitude of speed oscillations that occur with the application of the standard P&O method. This proposed method also shows better performance than the MEPT method used in previous studies in terms of accelerating maximum efficiency achievement.

244

Authors: Gokulapriya Raman, Ganesh Kumar Raj

Web usage behaviour mining is a substantial research problem to be resolved as it identifies different user’s behaviour pattern by analysing web log files. But, accuracy of finding the usage behaviour of users frequently accessed web patterns was limited and also it requires more time. Mutual Information Pre-processing based Broken-Stick Linear Regression (MIP-BSLR) technique is proposed for refining the performance of web user behaviour pattern mining with higher accuracy. Initially, web log files from Apache web log dataset and NASA dataset are considered as input. Then, Mutual Information based Pre-processing (MI-P) method is applied to compute mutual dependence between the two web patterns. Based on the computed value, web access patterns which relevant are taken for further processing and irrelevant patterns are removed. After that, Broken-Stick Linear Regression analysis (BLRA) is performed in MIP-BSLR for Web User Behaviour analysis. By applying the BLRA, the frequently visited web patterns are identified. With the identification of frequently visited web patterns, MIP-BSLR technique exactly predicts the usage behaviour of web users, and also increases the performance of web usage behaviour mining. Experimental evaluation of MIP-BSLR method is conducted on factors such as pattern mining accuracy, false positives, time requirements and space requirements with respect to number of web patterns. Outcomes show that the proposed technique improves the pattern mining accuracy by 14%, and reduces the false positive rate by 52%, time requirement by 19% and space complexity by 21% using Apache web log dataset as compared to conventional methods. Similarly, the pattern mining accuracy of NASA dataset is increased by 16% with the reduction of false positive rate by 47%, time requirement by 20% and space complexity by 22% as compared to conventional methods.

270

Authors: Abhijit Halkai, Sujatha Terdal

A sensor network operates wirelessly and transmits detected information to the base station. The sensor is a small sized device, it is battery-powered with some electrical components, and the protocols should operate efficiently in such least resource availability. Here, we propose a novel improved framework in large scale applications where the huge numbers of sensors are distributed over an area. The designed protocol will address the issues that arise during its communication and give a consistent seamless communication system. The process of reasoning and learning in cognitive sensors guarantees data delivery in the network. Localization in Scarce and dense sensor networks is achieved by efficient cluster head election and route selection which are indeed based on cognition, improved Particle Swarm Optimization, and improved Ant Colony Optimization algorithms. Factors such as mobility, use of sensor buffer, power management, and defects in channels have been identified and solutions are presented in this research to build an accurate path based on the network context. The achieved results in extensive simulation prove that the proposed scheme outperforms ESNA, NETCRP, and GAECH algorithms in terms of Delay, Network lifetime, Energy consumption.

296

Authors: Yossra H. Ali, Suhaila N. Mohammed

Malaria is a curative disease, with therapeutics available for patients, such as drugs that can prevent future malaria infections in countries vulnerable to malaria. Though, there is no effective malaria vaccine until now, although it is an interesting research area in medicine. Local descriptors of blood smear image are exploited in this paper to solve parasitized malaria infection detection problem. Swarm intelligence is used to separate the red blood cells from the background of the blood slide image in adaptive manner. After that, the effective corner points are detected and localized using Harris corner detection method. Two types of local descriptors are generated from the local regions of the effective corners which are Gabor based features and color based features. The extracted features are finally fed to Deep Belief Network (DBN) for classification purpose. Different tests were performed and different combinations of feature types are attempted. The achieved results showed that when using combined vectors of local descriptors, the system gives the desired accuracy which is 100%. The achieved result demonstrates the effectiveness of using local descriptors in solving malaria infection detection problem.

316

Authors: Eman K. Elsayed, Doaa R. Fathy

Dynamic Sign Language Recognition aims to recognize hand gestures of any person. Dynamic Sign Language Recognition systems have challenges in recognizing the semantic of hand gestures. These challenges come from the personal differences in hand signs from one person to another. Real-life video gesture frames couldn’t be treated as frame-level as a static sign. This research proposes a semantic translation system for dynamic hand gestures using deep learning and ontology. We used the proposed MSLO (Multi Sign Language Ontology) in the semantic translation step. Also, any user can retrain the system to be a personal one. We used Three-dimensional Convolutional Neural Networks followed by Convolutional long short-term memory to improve the recognition accuracy in Dynamic sign language recognition. We applied the proposed system on three dynamic gesture datasets from color videos. The recognition accuracy average was 97.4%. We did all the training and testing processes on the Graphics Processing Unit with the support of Google Colab. Using "Google Colab" in the training process decreases the average run time by about 87.9%. In addition to adding semantic in dynamic sign language translation, the proposed system achieves good results compared to some dynamic sign language recognition systems.

340

Authors: Ashraf Yahia Hassan, Ali Saleh

Zero-Forcing (ZF) detector is used in Space-Division Multiplexing (SDM) receiver to remove interference among the received symbols. Previous works showed that the power of channel noise is enhanced in the output of the ZF detector. They recommend using the ZF detector when the received Signal-to-Noise Ratio (SNR) is high. This work proves that the performance of the ZF detector depends on the eigenvalues of the channel correlation matrix. The paper shows that if the sum of the eigenvalues of this correlation matrix is equal to the rank of the channel matrix, the ZF detector will not enhance noise power at its outputs. Moreover, if the sum of the eigenvalues is smaller than the rank of the channel matrix, the ZF detector will reduce noise power at its outputs. In this work, a theorem, which demonstrates the performance of the ZF detector in SDM receiver, is introduced and proved. The proposed work uses smart antennas in the transmitter and receiver to control the elements and eigenvalues of the channel matrix. The introduced theorem and a complete SDM receiver with ZF detector are simulated and evaluated at different conditions with different criteria. A real-time SDM receiver with ZF detector is also implemented and evaluated. The simulation and implementation results are shown at the end of this study. The results of the proposed systems show that a ZF detector can be used to remove interference in the SDM system without enhancing the channel noise.

368

Authors: Rafid Abbas Ali, Asaad M. Abbas, Hazim G. Daway

Medical images are often adversely affected by a lack of clarity due to the limited representation of the color gamut. In this research, three types of medical images microscopic, magnetic resonance and x-ray images were enhanced by using a Fuzzy Logic by Stretch Membership Function (FLSMF). The Stretch Membership Function increased the dynamic range for the compounds red, green and blue in the medical images which have a few ranges. The FLSMF algorithm was compared with other methods by calculating the entropy value, wavelet quality evaluator and lightness order error. The analysis of the results showed that the proposed method succeeded in enhancing the contrast of the different types of medical images where it had high average values for the entropy (6.95) and wavelet quality evaluator (0.08), and a small average value for the lightness order error (60.77).

387

Authors: Noor Ifada, Richi Nayak

The tag-based recommendation systems that are built based on tensor models commonly suffer from the data sparsity problem. In recent years, various weighted-learning approaches have been proposed to tackle such a problem. The approaches can be categorized by how a weighting scheme is used for exploiting the data sparsity – like employing it to construct a weighted tensor used for weighing the tensor model during the learning process. In this paper, we propose a new weighted-learning approach for exploiting data sparsity in tag-based item recommendation system. We introduce a technique to represent the users’ tag preferences for leveraging the weighted-learning approach. The key idea of the proposed technique comes from the fact that users use different choices of tags to annotate the same item while the same tag may be used to annotate various items in tag-based systems. This points out that users’ tag usage likeliness is different and therefore their tag preferences are also different. We then present three novel weighting schemes that are varied in manners by how the ordinal weighting values are used for labelling the users’ tag preferences. As a result, three weighted tensors are generated based on each scheme. To implement the proposed schemes for generating item recommendations, we develop a novel weighted-learning method called as WRank (Weighted Rank). Our experiments show that considering the users' tag preferences in the tensor-based weighting-learning approach can solve the data sparsity problem as well as improve the quality of recommendation.

410

Authors: Mohammed Hussein Jabardi, Asaad Sabah Hadi

One of the most popular social media platforms, Twitter is used by millions of people to share information, broadcast tweets, and follow other users. Twitter is an open application programming interface and thus vulnerable to attack from fake accounts, which are primarily created for advertisement and marketing, defamation of an individual, consumer data acquisition, increase fake blog or website traffic, share disinformation, online fraud, and control. Fake accounts are harmful to both users and service providers, and thus recognizing and filtering out such content on social media is essential. This study presents a new approach to detect fake Twitter accounts using ontology and Semantic Web Rule Language (SWRL) rules. SWRL rules-based reasoner is utilized under predefined rules to infer whether the profile is trust or fake. This approach achieves a high detection accuracy of 97%. Furthermore, ontology classifier is an interpretable model that offers straightforward and human-interpretable decision rules.

433

Authors: Ramandeep Kaur, Sumit Mittal

Seamless integration of the new and existing networks is required to create next generation wireless networks capable of providing wide range of services coupled with uninterrupted connectivity and high speeds offering unprecedented quality of service to users. Smooth and proficient handoff techniques are important to achieve optimal network performance in heterogeneous wireless environments. This paper proposes Multiple Criteria Decision Making (MCDM) techniques of both non-fuzzy (Technique for Order of Performance by Similarity to Ideal Solution (TOPSIS)) and fuzzy (Fuzzy TOPSIS (FTOPSIS)) type for network ranking. The proposed algorithm uses priority ranking to assign weights to six parameters by using Fuzzy Analytic Hierarchy Process (FAHP). Further, TOPSIS and FTOPSIS techniques are compared for various traffic classes and used for network selection in the prevalent network conditions. The results show that contingent on the traffic classes, the FTOPSIS selects the ideal network with diminished ping-pong rate and call blocking probability as compared to TOPSIS while accounting for the uncertain environment.

453

Authors: Abdul Syukur, Deden Istiawan

LQ45 is an Indonesia Stock Exchange Index (ISX) incorporate of 45 companies that meet certain criteria to target investors for selecting certain stocks. The prediction of stock price direction in the financial world is a major issue. The implementation of machine learning and other algorithms for market price analysis and forecasting is a very promising field. Different types of classification algorithms were used to predict the stock market. However, when individual studies are considered separately there is no clear consensus that algorithms work best. In this research, a comparison framework is proposed, which aims to benchmark the performance of a wide range of classification models and use them to predict the LQ45 index. The data in this research contains the transaction level and capitalization size are obtained from the Indonesian Stock Exchange (ISX). For analysis purposes, we set out 10 classifiers that can be used to build classification models and test their performance in the LQ45 dataset. The performance criterion chosen to measure this effect is accuracy, recall, and precision. The results showed that the random forest algorithm had the best performance for predicting the LQ45 index. Whilst the classification and regression trees, C4.5, support vector machine, and logistic regression algorithms also perform well. Besides, the models based on traditional statistical-based learners that are Naïve Bayes and linear discriminant analysis seem to underperform for predicting the LQ45 index. These results are not only beneficial to enrichment the machine learning techniques literature but also have a significant influence on the stock market prediction in terms of the ability to predict the LQ45 index.

474

Authors: Farah Afianti, Cahya Asrini, Dedy Rahman Wijaya

Wireless sensor nodes help people to monitor any operational mechanism. It is important to be implemented in industrial automation in detecting unauthorized activity or malicious sensor nodes by the outsider. However, a massive number of devices or things which are needed to be controlled become a challenge. A fast mechanism to recognize a member or a valid device is needed to minimize production costs and avoid future procedural errors. It becomes an open problem for the implementation of massive scale Wireless Sensor Networks (WSN). Hence, a scalable two-dimensional Bloom filter is proposed in this paper to deal with this issue. The low computation complexity and the storage efficiency requirements by the two-dimensional Bloom filter are proven by the experimental results. The insertion, deletion and query procedure of the proposed method only need O (1), while the storage usage of two-dimensional Bloom filters not only lower than counting Bloom filter in average 131 bits difference but also approaches accommodative Bloom filter in about 10 bits difference. These two parameters support a fast membership scheme. Furthermore, the analysis of value initialization has been performed. To the best of our knowledge, the investigation of this parameter has not been addressed by other studies. This process aims to get the best scenario to increase scalability. Several recommendations for selecting dimension number has been stated which is useful for efficient storage usage and reduce false positive.

496

Authors: Reza Fauzan, Daniel Siahaan, Siti Rochimah, Evi Triandini

The use case diagram is one of the diagrams commonly taught in colleges of computer science. Assessment of use case diagrams is often an obstacle for a teacher in the learning process. It is due to the interpersonal and intrapersonal problems of the teacher in assessing. Interpersonal problems are caused by the absence of an assessment standard among teachers. Intrapersonal problems are caused by the inconsistency of a teacher in assessing many diagrams of student answers. This research aims to create a semantic use case diagram automatic assessment method. Semantic assessment is divided into two kinds, namely property and relationship. All information used is a label translated from the XMI document. Similarity assessment between labels used cosine similarity, employing WuPalmer to perform WordNet searches. The results showed that the proposed method had a substantial agreement with the teacher as an expert; however, a teacher tends to look at property information rather than relationship information to assess use case diagrams.

516

Authors: Ana Alimatus Zaqiyah, Diana Purwitasari, Chastine Fatichah

Spam detection frequently categorizes product reviews as spam and non-spam. The spam reviews may contain texts of fake reviews and non-review statements describing unrelated things about products. Most of the publicly available spam reviews are labelled as fake reviews, while non-spam texts that are not fake reviews could contain non-review statements. It is crucial to notice those non-review statements since they convey misperception to consumers. Non-review statements are hardly found, and those statements of large and long texts often need to be manually labelled, which is time-consuming. Because of the rareness in finding non-review statements, there is an imbalanced condition between non-spam as a major class and spam that consists of the non-review statement as a minor class. Augmenting fake reviews to add spam texts is ineffective because they have similar content to non-spam such as some opinion words of product features. Thus, the text generation of non-review statements is preferable for adding spam texts. Some text generation issues are the frequent neural network-based methods require much learning data, and the existing pre-trained models produce texts with different contexts to non-review statements. The augmented texts should have similar content and context represented by the structure of the non-review statement. Therefore, we propose a text generation model with content and structure-based preprocessing to produce non-review statements, which is expected to overcome imbalanced data and give better spam detection results in product reviews. Structure-based preprocessing identifies the feature structures of non-opinion words from part-of-speech tags. Those features represent the context of spam reviews in unlabeled texts. Then, content-based preprocessing appoints selected topic modeling results of non-review statements from fake reviews. Our experiments resulted an improvement on the metric value of ± 0.04, called as BLEU (Bi-Lingual Evaluation Understudy) score, for the correspondence evaluation between generated and trained texts. The metric value indicates that the generated texts are not quite identical to the trained texts of non-review statements. However, those additional texts combined with the original spam texts gave better spam detection results with an increasing value of more than 40% on average recall score.

541

Authors: Cahya Rahmad, Kohei Arai, Rosa A. Asmara, Ekojono Ekojono, Dimas R. H. Putra

Face recognition plays an important role in the identity recognition system, the color and geometry feature has been claimed able to be used as parameter for face recognition. This study aims to analize the performance of geometric features, color features, and both of them on the human face using Gaussian Naïve Bayes (GNB) and the other Machine Learning method. This study using various geometric features: the distance between the eyes, nose, mouth by using Euclidean distance, and classified using GNB, K-Nearest Neighbour (KNN), and Support Vector Machine (SVM). The result compared with color feature: normalized RGB values, mean of normalized RGB, and RGB Variant as color features. The feature values obtained are assembled and processed using GNB and the other ML method to classified and recognized the faces. The dataset obtained from Aberdeen faces the dataset, which has 687 color faces from Ian Craw at Aberdeen. Between 1 and 18 images of 90 individuals. Some variations in lighting, varied viewpoints, and the resolution have varied between 336x480 to 624x544. The experimental results show that the system successfully recognized the face based on the determined algorithm and based on three models, SVM reached nearly 74.83%, GNB reached nearly 74.67%, and KNN with K = 5 reached nearly 72.17%.