With the ever increasing of internet applications and social networking sites, people nowadays can easily express their feelings towards any products and services. These online reviews act as an important source for further analysis and improved decision making. These reviews are mostly unstructured by nature and thus, need processing like sentiment analysis and classification to provide a meaningful information for future uses. In text analysis tasks, the appropriate selection of words/features will have a huge impact on the effectiveness of the classifier. Thus, this paper explores the effect of the pre-processing strategies in the sentiment analysis of online movie reviews. In this paper, supervised machine learning method was used to classify the reviews. The support vector machine (SVM) with linear and non-linear kernel has been considered as classifier for the classification of the reviews. The performance of the classifier is critically examined based on the results of precision, recall, f-measure, and accuracy. Two different features representations were used which are term frequency and term frequency-inverse document frequency. Results show that the pre-processing strategies give a significant impact on the classification process.

The discovery of an active feature extraction technique has been the focus of many researchers to improve the performance of classification methods, such as for sentiment analysis. Many of them have shown interest in using word embeddings especially Word2Vec as the features for text classification tasks. Its ability to model high-quality distributional semantics among words has contributed to its success in many of the functions. Despite the success, Word2Vec features are high dimensional that lead to an increase in the complexity of the classifier. In this paper, an effective method for feature extraction based on Word2Vec is proposed for sentiment analysis. The process discovers polarity clusters of the terms in the vocabulary through Word2Vec and opinion lexical dictionary. The features vector for each text is constructed from the polarity clusters, which lead to a lower-dimensional vector to represent the text. This paper also investigates the effect of two opinion lexical dictionaries on the performance of sentiment analysis, and one of the dictionaries are created based on SentiWordNet. The effectiveness of the proposed method is evaluated on the IMDB with two classifiers, namely the Logistic Regression and the Support Vector Machine. The result is promising, showing that the proposed method can be more effective than the baseline approaches.

CLASSIFICATION MODEL FOR HOTSPOT OCCURRENCES USING SPATIAL DECISION TREE ALGORITHM

With the technological development and means of communication, the Internet of Things (IoT) has become an essential role in providing many services in daily life through millions of heterogeneous but interconnected devices and nodes. This development is opening to many security and privacy challenges that can cause complete network breakdown, bypassed access control or the loss of critical data. This paper attempts to provide a preliminary analysis for malware detection within data generated by IoT-based devices and services in the form of operational codes (Opcode) sequences. Three machine learning algorithms are evaluated and compared for accuracy, precision, recall and F-measure. The results showed that the Random Forest (RF) achieved the best accuracy of 98%, followed by SVM and k-NN, both with 91%. The results are further analyzed based on the Receiver Operating Characteristic (ROC) curve and Precision-Recall curve to further illustrate the difference in performance of all three algorithms when dealing with IoT data.

Predicting hotspot occurrence as an indicator of forest and land fires is essential in developing an early warning system for fire prevention. This work applied a spatial decision tree algorithm on spatial data of forest fires. The algorithm is the improvement of the conventional decision tree algorithm in which the distance and topological relationships are included to grow up spatial decision trees. Spatial data consist of a target layer and ten explanatory layers representing physical, weather, socio-economic and peatland characteristics in the study area Rokan Hilir District, Indonesia. Target objects are hotspots of 2008 and non-hotspot points. The result is a pruned spatial decision tree with 122 leaves and the accuracy of 71.66%. The spatial tree has produces higher accuracy than the non-spatial trees that were created using the ID3 and C4.5 algorithm. The ID3 decision tree has accuracy of 49.02% while the accuracy of C4.5 decision tree reaches 65.24%.