Andrei D. Coronel

eHealth TABLET: A Developing Country Perspective in Managing the Development and Deployment of a Mobile - Cloud Electronic Medical Record for Local Government Units

In January 2013, the eHealth TABLET (Technology Assisted Boards for Local government unit Efficiency and Transparency) project began with a two-fold objective of: 1) creating a tablet based system that will integrate existing health information systems to address the national objective of a unified health information management system by 2015 and 2) to create a transparency layer at the local government units such that communication lines between municipal health officers and the mayor are monitored. Bottom up approach was used to ensure that all features requested by multi-stakeholders are included in the design of the system. The end product was a mobile - web based system with the mobile application having three main components: the electronic medical record (EMR) application which comprises of the patient record and diagnosis module, the requests/approval application, and the dashboard application for data visualization. Responding to the needs of intended users, the web based application provides the following features: web auxiallry entry, aggregated disease report application and usage monitoring. Regular usage monitoring increased usage over time. For ICT development projects in public health, iteratve involvement of multi-stakeholders is necessary to ensure higher acceptance and adoption. From a design perspective, technologies should be designed to be interoperable such that interfacing with existing systems will be seamless.

Bottom Up Approach and Devolved Design of a Health Information System: eHealth TABLET

Health care is expensive in the Philippines because of the lack of medical experts and facilities that are able to reach remote areas in the country. At the same time, access to real time health information is also undermined by several layers of paper based data entry. In areas where there are existing information systems, the burden is placed on the health worker in using several information systems to address various health concerns. This paper presents eHealth TABLET (Technology Assisted Boards for Local government unit Efficiency and Transparency), a local mobile (tablet-based) electronic medical record system and dashboard for decision making (coupled with a Doctor-Mayor communication feature) designed to answer problems in accessibility, efficiency and transparency following a bottom up approach and devolved approach in designing the system. As a local Electronic Medical Record (EMR) system, it provides the municipalities with a tailor-fit simple patient record system to better address the needs of their patients. As a health dashboard, it provides accurate and real-time visualizations of local patient data for decision-making purposes. As a messaging system, it provides a more efficient and transparent communication system between the Mayor (Local Chief Executive) and the Doctor (Health Officer).

Towards building a predictive model for remote river quality monitoring for mining sites

Most, if not all, mining sites in the Philippines are not equipped with expensive or modern monitoring tools to check for quality of soil, water and air elements which are relevant to ensure safety and wellness of miners. This study focused on the development of low cost mobile electronic sensors to monitor quality of water from rivers near mining sites. Low cost electronic sensors connected to a smart phone were developed to capture dissolved oxygen (DO2), pH, Turbidity, Temperature, and Salinity. The data for mercury (Hg) and arsenic (As) were obtained through AAS analyses to form baseline data for the model. Data was collected for over a period of one year, with site visits once every two months. A conditional inference tree (ctree) using recursive binary partitioning was used to generate the prediction model using 70 - 30 split on the training and test data set. The multi-feature model returns Good, Not Good or Unknown based on the scores of each element. The results showed a possible three feature model with significant results for site, salinity and pH balance.

Building an Initial Fitness Function Based on an Identified Melodic Feature Set for Classical and Non-Classical Melody Classification

Algorithmic Composition for music is a progressive field of study. The success of an automated music- generating algorithm, however, depends heavily on the fitness function that is used to score the generated music. Artificial intelligence in this context of music scoring can use a fitness function that is generally based on music features that a given algorithm is programmed to measure. This study explores the features that are important for melody generation by investigating those that can separate classical from non-classical music in the context of melody. The jSymbolic tool was used to collect 160 standard features from 400 music files. Using C4.5 algorithm to select significant features used in a classical vs. non-classical melody classification challenge, and then performing a comparison with a suggested feature set by running Naïve-Bayes and SVM classifiers, the study was able to determine a candidate set of melodic features that can be used for building an initial fitness function that separates classical from non-classical melodies with high accuracy as revealed by SVM ten-fold cross validation.

A hybrid approach towards improved artificial neural network training for short-term load forecasting

The power of artificial neural networks to form predictive models for phenomenon that exhibit non-linear relationships is a given fact. Despite this advantage, artificial neural networks are known to suffer drawbacks such as long training times and computational intensity. The researchers propose a two-tiered approach to enhance the learning performance of artificial neural networks for phenomenon with time series where data exhibits predictable changes that occur every calendar year. This paper focuses on the initial results of the first phase of the proposed algorithm which incorporates clustering and classification prior to application of the backpropagation algorithm. The 2016--2017 zonal load data of France is used as the data set. K-means is chosen as the clustering algorithm and a comparison is made between Naïve Bayes and k-Nearest Neighbors to determine the better classifier for this data set. The initial results show that electrical load behavior is not necessarily reflective of calendar clustering even without using the min-max temperature recorded during the inclusive months. Simulating the day-type classification process using one cluster, initial results show that the k-nearest neighbors outperforms the Naïve Bayes classifier for this data set and that the best feature to be used for classification into day type is the daily min-max load. These classified load data is expected to reduce training time and improve the overall performance of short-term load demand predictive models in a future paper.

Towards Automated Counter-Melody Generation for Monophonic Melodies

Algorithmic composition has focused on creating music from algorithms, stemming from the capacity to convert notes into numbers and vice versa, thus allowing simple to complex algorithmic manipulations. The focus of these studies has either been the creation of melodies, chords, accompaniments or entire songs. This study focuses on a relatively underexplored topic on the algorithmic generation of a counter-melody from a given melody. Using a method based on existing knowledge of generating chords and music theory on compatible notes and chord progressions, combined with concepts of machine learning and tree traversal techniques for generating chords, this study was able to generate 200 counter-melodies from 100 inputs, involving two generation techniques per input. The results show that counter-melodies were successfully generated based on chord progression generation and note selection approaches, and after subjecting the counter-melodies to proper subject evaluation, the average scores of 2.89 and 3.02 on a 5-point evaluation criteria reveal that the counter-melodies are musically fit for the original melodies they were based from.

Comparative analysis of tree classification models for detecting fusarium oxysporum f. sp cubense (TR4) based on multi soil sensor parameters

Use of wireless sensor networks and smartphone integration design to monitor environmental parameters surrounding plantations is made possible because of readily available and affordable sensors. Providing low cost monitoring devices would be beneficial, especially to small farm owners, in a developing country like the Philippines, where agriculture covers a significant amount of the labor market. This study discusses the integration of wireless soil sensor devices and smartphones to create an application that will use multidimensional analysis to detect the presence or absence of plant disease. Specifically, soil sensors are designed to collect soil quality parameters in a sink node from which the smartphone collects data from via Bluetooth. Given these, there is a need to develop a classification model on the mobile phone that will report infection status of a soil. Though tree classification is the most appropriate approach for continuous parameter-based datasets, there is a need to determine whether tree models will result to coherent results or not. Soil sensor data that resides on the phone is modeled using several variations of decision tree, namely: decision tree (DT), best-fit (BF) decision tree, functional tree (FT), Naive Bayes (NB) decision tree, J48, J48graft and LAD tree, where decision tree approaches the problem by considering all sensor nodes as one. Results show that there are significant differences among soil sensor parameters indicating that there are variances in scores between the infected and uninfected sites. Furthermore, analysis of variance in accuracy, recall, precision and F1 measure scores from tree classification models homogeneity among NBTree, J48graft and J48 tree classification models.

Streamlining machine learning in mobile devices for remote sensing

Mobile devices have been at the forefront of Intelligent Farming because of its ubiquitous nature. Applications on precision farming have been developed on smartphones to allow small farms to monitor environmental parameters surrounding crops. Mobile devices are used for most of these applications, collecting data to be sent to the cloud for storage, analysis, modeling and visualization. However, with the issue of weak and intermittent connectivity in geographically challenged areas of the Philippines, the solution is to provide analysis on the phone itself. Given this, the farmer gets a real time response after data submission. Though Machine Learning is promising, hardware constraints in mobile devices limit the computational capabilities, making model development on the phone restricted and challenging. This study discusses the development of a Machine Learning based mobile application using OpenCV libraries. The objective is to enable the detection of Fusarium oxysporum cubense (Foc) in juvenile and asymptomatic bananas using images of plant parts and microscopic samples as input. Image datasets of attached, unattached, dorsal, and ventral views of leaves were acquired through sampling protocols. Images of raw and stained specimens from soil surrounding the plant, and sap from the plant resulted to stained and unstained samples respectively. Segmentation and feature extraction techniques were applied to all images. Initial findings show no significant differences among the different feature extraction techniques. For differentiating infected from non-infected leaves, KNN yields highest average accuracy, as opposed to Naive Bayes and SVM. For microscopic images using MSER feature extraction, KNN has been tested as having a better accuracy than SVM or Naive-Bayes.

Evolutionary algorithm-based composition of hybrid-genre melodies using selected feature sets

Algorithmically generating music using specialized algorithms is a growing focus in computer science. The success of these specialized algorithms in generating music, however, depends heavily on the fitness function that is used to score the generated music and equally as important is how the fitness function is designed. Artificial intelligence in the computational composition can use certain feature set values derived from melodic analysis to serve as criteria for these fitness functions. This study explores two methods in defining the key features to be used as fitness criteria for algorithmic music generation of music that can be considered under a mix of two musical genres or hybrid-genre music. The jSymbolic tool was used to extract 101 features from musical pieces that fall under two genres. This was then reduced to a smaller feature set for use as fitness criteria. Two methods for feature reduction was explored; a decision-tree-based technique and a high-correlation-filtering technique. The study was able to confirm that each technique can be used to compose hybrid-genre music with 86% success-rate as confirmed by SVM when validated under the same dataset used in the study. This study does not claim to consistently result in a high success rate for all existing datasets.

Developing a Philippine Cancer Grid. Part 1: Building a Prototype for a Data Retrieval System for Breast Cancer Research Using Medical Ontologies

Cancer is a leading cause of morbidity and mortality in the Philippines. Developed within the context of a Philippine Cancer Grid, the present study used web development technologies such as PHP, MySQL, and Apache server to build a prototype data retrieval system for breast cancer research that incorporates medical ontologies from the Unified Medical Language System (UMLS).