Search

Software-defined networking (SDN) paradigm leads to innovation and simplicity in network configuration and management. Wireless sensor networks (WSNs) are becoming an essential component of the Internet of Things (IoTs). The combination of these two models gives rise to a new paradigm: Software-defined wireless sensor networks (SDWSNs). This paper provides a systematic review on SDWSNs. The study also discusses open challenges of further research on SDWSNs.

Batik is a method of cloth decoration by dyeing process which is practiced particularly in Java, Indonesia. The batik pattern varies from one region to another. In this paper, we explore and compare the shallow and deep learning model to classify Lasem batik motifs automatically. Several shallow models are employed, such as the support vector machines, Gaussian naive Bayes, and decision tree with the gray level co-occurrence matrix as the features. For the deep learning models, we use the convolutional neural network and deep belief network. We collect 698 batik images which can be classified into seven different types of Lasem batik motif. From the experiments, we found that the shallow model, particularly the support vector machines with linear function kernel performs best, even compared to the deep learning models.

We proposed unsupervised learning, the Intelligent K-Medoids Algorithm to predict, the length of a study time of universitys students. This algorithm automatically clusters all students based on their 25 weighted scores from 25 different subject as the features. We tested the implementation of the algorithm using 240 students scores. These 240 students have graduated and their graduation time is considered for labeling the cluster. The result is 7 clusters with silhouette value of 0.2416. Each cluster is labeled according to the range of student graduation time. The range in each cluster exists due to the existence of students whose majority of scores are similar, but their graduation times are different. Academic leaving or extending the completion of thesis are the other factors contributing the range graduation time in each cluster. The prediction by k-folding 240 data to 5 subsets results average prediction accuracy of 99.58.

This paper proposes a reliable minimum vector variance regression algorithm for robust supervised impervious mapping. The mapping is done with a conventional two phase process; training and mapping process. The outcome of training process is the robust regression models useful for the knowledge base of mapping land cover. The robust regression model is built from the existing robust depth minimum vector variance subsample. The case of research is a metropolitan area consisting of megacities surrounding Jakarta, Jabodetabek. The urban population in the Jabodetabek area is very high. The urbanization is closely related to the percentage of impervious area and indicates the quality of the environment. The evaluation mapping provides that the robust depth minimum vector variance regression is an effective method for the impervious land cover mapping of Jabodetabek. Keywords: depth function, impervious, minimum vector variance, robust

The pattern of signature and handwriting are unique, so they can be utilised as an authentication system. This research proposed a method of signature and handwriting recognition on a mobile device using the Gray Level Co-occurrence Matrix (GLCM) for texture-based feature extraction and the bootstrap for performing single classifier model. The proposed method is successfully implemented in the offline and online application. The offline experiment of signature and handwriting from the same user produces accuracy 100%. In a cross evaluation using different users as model and target, the experiment performs accuracy around 34% and 44% for signature and handwriting data, respectively. In the case study of the training and testing data from the same user on mobile devices, the experiment using stylus and finger produces accuracy 84.62% and 88.46%, respectively for online signature recognition, and 70% and 90% for online handwriting recognition.

Wireless sensor networks with battery-operated nodes have the limitation in the energy availability at each node. One visible solution is to add the nodes with energy harvesting capability. By harnessing energy from the environment or other sources, the nodes can prolong its operation. This paper provides a brief review of energy harvesting systems - protocol, energy sources, and storage technologies. The study also discusses some opportunities offurther research on energy harvesting networks.

Perancangan aplikasi pendeteksian tingkat kesamaan antar dokumen ini dimaksudkan untuk menyimpan data tesis mahasiswa dan aplikasi dapat digunakan untuk membandingkan data skripsi mahasiswa dengan data baru yang ada. Hasil dari proses deteksi adalah persentase tingkat kesamaan. Algoritma yang digunakan dalam merancang aplikasi ini adalah Algoritma Winnowing. Algoritma ini termasuk dalam salah satu metode fingerprinting dokumen. Algoritma ini menggunakan rolling hashing untuk melakukan proses hashing dan menggunakan koefisien jaccard untuk menghitung tingkat kemiripan. Dalam pengujian aplikasi ini dengan data dummy diketahui bahwa aplikasi ini sangat bergantung pada urutan masing-masing lokasi sub string. Aplikasi ini juga telah diuji untuk membandingkan sejumlah data skripsi yang ada dengan persentase tingkat kesamaan di bawah 30 persen. Aplikasi ini tidak bisa menentukan apakah sebuah dokumen plagiat atau tidak, namun aplikasi ini bisa memberikan informasi berupa persentase tingkat kesamaan dan persentase dapat digunakan oleh pihak yang memiliki wewenang untuk menentukan.

Aplikasi prediksi masa studi mahasiswa merupakan aplikasi yang mengimplementasikan metode Voting Feature Interval 5 pada fitur prediksi masa studi mahasiswa. Hasil prediksi masa studi akan digunakan sebagai masukkan bagi dosen penasehat akademik dan ketua program studi untuk menyusun rencana studi yang tepat bagi mahasiswa. Studi kasus yang digunakan dalam aplikasi adalah prediksi masa studi mahasiswa Program Studi Teknik Informatika, di salah satu perguruan tinggi swasta di Jakarta, angkatan 2013 dan 2014. Adapun data pembelajaran yang digunakan adalah data nilai mahasiswa, program studi Teknik Informatika di perguruan tinggi swasta tersebut angkatan 2008 sampai 2012. Metode Voting Feature Interval 5 bekerja dengan mengklasifikasikan data nilai mahasiswa berdasarkan nilai vote yang terbentuk dari hasil pembelajaran data latih. Hasil pengujian aplikasi menunjukkan bahwa aplikasi berhasil menghasilkan prediksi masa studi bagi mahasiswa dengan tingkat akurasi yaitu sebesar 73,33%.

University Course Timetabling Problem (UCTP) belongs to Constraint Satisfaction Problems (CSPs), which are the set of objects whose state must satisfy a number of constraints. The constraints, in this case, are related to characteristics and regulations of a particular university. Certainly, these will vary from one university to the other. A number of approaches have provided feasible and optimal solutions for UCTP. However, their solutions are still based certain university's constraints. Our approach has given another way to occupy various constraints for various universities. Dynamic Constraint Matching (DCM) consists of constraints logical formulation, collision matrix generation, and validation using the collision matrix. Our experiment, using 93 subjects offered in Faculty of Information Technology Tarumanagara University, has shown that all constraints, taken from the characteristics and regulations of the Faculty, can be formulated successfully. When DCM were integrated with Vertex Graph Coloring (VGC) as one of the guaranteed optimal solutions for UCTP, the approach results a course schedule that does not contain any hard or soft constraints violations. The processing time can be said fast, which is less than 1 minutes.

Crescenzi et al. have proposed a fast full-expansion-compression (FEC) technique that requires exactly three memory accesses for each IP lookup. However, prefix updates on their FEC scheme require the forwarding table data structure to be rebuilt from scratch. Therefore, fast table reconstruction is important to make the scheme effective for use in a dynamic environment. In this paper, we propose an efficient technique to reduce the FEC table construction time. Our approach generates the RLE sequences for the forwarding table directly from a list of decreasing lexicographic ordered prefixes to avoid the expansion step of the previous technique. We also propose an improved unification technique to construct the FEC tables more efficiently. Our approach has been implemented in C, and several experiments using six databases from some real IPv4 router show that our proposed technique runs 4 to 29 times faster than the previous technique

We propose a dynamic-full-expansion-compression (DFEC) technique that requires exactly three memory accesses for each IP address lookup as that required in the static-FEC. A prefix update in the static scheme needs table reconstruction that requires hundreds of milliseconds, and hence the approach is unsuitable for use in a dynamic-environment. DFEC selectively decompresses the table to obtain addresses that are affected by a prefix update, modifies their next-hops, and recompresses the data structure. The resulting tables will be identical to what would be obtained by a full reconstruction, but at much lower computational cost. We have implemented our algorithm in ANSI-C and have run it on a 3.2 GHz Pentium-4 to build an FEC table of each of the following five databases: AADS2001, Mae- West2001, Mae-East 1997, Paix2001, and PB2001. Our simulations show that an update in DFEC on each of the tables takes less than 11 microseconds.

Intelligent Kernel K-Means is a fully unsupervised clustering algorithm based on kernel. It is able to cluster kernel matrix without any information regarding to the number of required clusters. Our experiment using gene expression of human colorectal carcinoma had shown that the genes were grouped into three clusters. Global silhouette value and davies-bouldin index of the resulted clusters indicated that they are trustworthy and compact. To analyze the relationship between the clustered genes and phenotypes of clinical data, we performed correlation (CR) between each of three phenotypes (distant metastasis, cancer and normal tissues, and lymph node) with genes in each cluster of original dataset and permuted dataset. The result of the correlation had shown that Cluster 1 and Cluster 2 of original dataset had significantly higher CR than that of the permuted dataset. Among the three clusters, Cluster 3 contained smallest number of genes, but 16 out of 21 genes in that cluster were genes listed in Tumor Classifier List (TCL).