Data Mining Practical Machine Learning Tools and Techniques

An overview of the self-organizing map algorithm, on which the papers in this issue are based, is presented in this article.

The Self-Organizing Map

『Google Scholar』の利点

An enhanced extreme learning machine model for river flow forecasting: State-of-the-art, practical applications in water resource engineering area and future research direction

A datatype with increasing importance in GIS is what we call the location history–a record of an entity’s location in geographical space over an interval of time. This paper proposes a number of rigorously defined data structures and algorithms for analyzing and generating location histories. Stays are instances where a subject has spent some time at a single location, and destinations are clusters of stays. Using stays and destinations, we then propose two methods for modeling location histories probabilistically. Experiments show the value of these data structures, as well as the possible applications of probabilistic models of location histories.

Project lachesis: Parsing and modeling location histories

Cervical cancer is the fourth most prevalent disease in women. Accurate and timely cancer detection can save lives. Automatic and reliable cervical cancer detection methods can be devised through the accurate segmentation and classification of Pap smear cell images. This paper presents an approach to whole cervical cell segmentation using a mask regional convolutional neural network (Mask R-CNN) and classifies this using a smaller Visual Geometry Group-like Network (VGG-like Net). ResNet10 is used to make full use of spatial information and prior knowledge as the backbone of the Mask R-CNN. We evaluate our proposed method on the Herlev Pap Smear dataset. In the segmentation phase, when Mask R-CNN is applied on the whole cell, it outperforms the previous segmentation method in precision (0.92±0.06), recall (0.91±0.05) and ZSI (0.91±0.04). In the classification phase, VGG-like Net is applied on the whole segmented cell and yields a sensitivity score of more than 96% with low standard deviation (±2.8%) for the binary classification problem and yields a higher result of more than 95% with low standard deviation (maximum 4.2% in accuracy measurement) for the 7-class problem in terms of sensitivity, specificity, accuracy, h-mean, and F1 score.

/pdf/segmentation-and-classification-of-cervical-cells-using-deep-16kaz9z6d5.pdf

Segmentation and classification of cervical cells using deep learning

Nowadays, many mobile phones have been equipped with sensors to enable the delivery of advanced features/services to the users. Accelerometer is one of the sensors that embedded to several types of mobile phones. Our earlier research has shown that data from mobile-phone embedded accelerometer can be used for activity recognition purpose [1]. As a continuation of the research towards the search for a suitable and reliable algorithm for real-time activity recognition using mobile phone, an evaluation and comparison study of the performance of seven different categories of classifier algorithms in classifying user activities were conducted. Five basic human activities (jogging, jumping, sitting, standing, and walking) were tested. The training and testing data were done using Weka 3.6.6 data mining tool. The overall accuracy rate for classifier training managed to exceed 96% and exceeded 90% for classifier testing, which are very encouraging results.

A Comparison Study of Classifier Algorithms for Mobile-phone's Accelerometer Based Activity Recognition

Face recognition can be considered one of the most successful biometric identification methods among several types of biometric identification including fingerprints, DNA, palm print, hand geometry, iris recognition, retina and odour/scent. Face recognition provides biometric identification that utilizes the uniqueness of faces for security purposes. The problem with face recognition using biometric identification is its lengthy process and the accuracy of the results. This paper proposes solutions for a faster face recognition process with accurate results. The proposed face recognition process was done using a hybrid process of Haar Cascades and Eigenface methods, which can detect multiple faces (55 faces) in a single detection process. This improved face recognition approach was able to recognize multiple faces with 91.67% accuracy level.

Multi-Faces Recognition Process Using Haar Cascades and Eigenface Methods

Patients with liver disease continue to increase and the symptoms of the disease is difficult to detect. Therefore many people suffer from liver damage but they feel healthy, it causes many medical practitioners to often fail to detect the disease. Failure to detect can mislead to improper medication and medical treatment. Therefore accurate detection is necessary to help the medical practitioner to give proper medication and medical treatment. Some researchers have been using data mining techniques to classify liver disease. The problem is, it is not easy to have the same consensus for a better algorithm in classifying liver disease. This study aims to identify if the patients have the liver disease based on the 10 important attributes of liver disease using a Decision Tree, Naive Bayes, and NBTree algorithms. The result shows NBTree algorithm has the highest accuracy, however the Naive Bayes algorithm gives the fastest computation time. This study presents promising results in giving recommendation if the patients have the disease.

Data Mining Techniques for Optimization of Liver Disease Classification

Activity recognition refers to the ability of a machine/device to recognize the activity of users. This area of research has attracted many works especially related to the context aware and ubiquitous computing area. Wearable accelerometers have been explored for this activity recognition purpose; however the impracticality of attaching accelerometers to the user presents significant issues. Accelerometers today are embedded in many mobile devices. This paper explores the potential and possibility of using these accelerometer data to determine user activity recognition. The initial experiments show encouraging results with a very good accuracy rate of 92%. A simple prototype developed supports the implementation of the recognition process conducted.

Teddy Mantoro

Papers

Segmentation and classification of cervical cells using deep learning

A Comparison Study of Classifier Algorithms for Mobile-phone's Accelerometer Based Activity Recognition

Multi-Faces Recognition Process Using Haar Cascades and Eigenface Methods

Data Mining Techniques for Optimization of Liver Disease Classification

Recognizing user activity based on accelerometer data from a mobile phone