Imas Sukaesih Sitanggang

Sequential pattern mining on library transaction data

Application of data mining techniques in library data results interesting and useful patterns that can be used to improve services in university libraries. This paper presents results of the work in applying the sequential pattern mining algorithm namely AprioriAll on a library transaction dataset. Frequent sequential patterns containing book sequences borrowed by students are generated for minimum supports 0.3, 0.2, 0.15 and 0.1. These patterns can help library in providing book recommendation to students, conducting book procurement based on readers need, as well as managing books layout.

Classification model for hotspot occurrences using spatial decision tree algorithm

Developing a predictive model for forest fires occurrence is an important activity in a fire prevention program. The model describes characteristics of areas where fires occur based on past fires data. It is essential as an early warning system for preventing forest fires, thus major damages because of fires can be avoided. This study describes the application of data mining technique namely decision tree on forest fires data. We improved the ID3 decision tree algorithm such that it can be utilized on spatial data in order to develop a classification model for hotspots occurrence. The ID3 algorithm which is originally designed for a non-spatial dataset has been improved to construct a spatial decision tree from a spatial dataset containing discrete features (points, lines and polygons). As the ID3 algorithm that uses information gain in the attribute selection, the proposed algorithm uses spatial information gain to choose the best splitting layer from a set of explanatory layers. The new formula for spatial information gain is proposed using spatial measures for point, line and polygon features. The proposed algorithm has been applied on the forest fire dataset for Rokan Hilir district in Riau Province in Indonesia. The dataset contains physical data, socio-economic, weather data as well as hotspots and non-hotspots occurrence as target objects. The result is a spatial decision tree with 276 leaves with distance from target objects to the nearest river as the first test layer and the accuracy on the training set of 87.69%. Empirical result demonstrates that the proposed algorithm can be used to join two spatial objects in constructing a spatial decision tree from a spatial dataset. The algorithm results a predictive model for hotspots occurrence from the real dataset on forest fires with high accuracy on the training set.

Classification model for hotspot occurrences using a decision tree method

Forest fires in Indonesia mostly occur because of errors or bad intentions. This work demonstrates the application of a decision tree algorithm, namely the C4.5 algorithm, to develop a classification model from forest fire data in the Rokan Hilir district, Indonesia. The classification model used is a collection of IF-THEN rules that can be used to predict hotspot occurrences for forest fires. The spatial data consist of the location of hotspot occurrences and human activity factors including the location of city centres, road and river networks as well as land cover types. The results were a decision tree containing 18 leaves and 26 nodes with an accuracy of 63.17%. Each leaf node holds positive and negative examples of hotspot occurrences whereas the root and internal nodes contain attribute test conditions: the distance from the location of examples to the nearest road, river, city centre and the land cover types for the area where the examples are located. Positive examples are hotspot locations in the study area and negative are randomly generated points within the area at least 1 km away from any positive example. The classification model categorized whether the region was susceptible to hotspots occurrences or not. The model can be used to predict hotspot occurrences in new locations for fire prediction.

Global and collective outliers detection on hotspot data as forest fires indicator in Riau Province, Indonesia

Forest fire in Indonesia is considered as an annual event that causes serious problems in health and environment especially in Sumatera and Kalimantan Islands. Studies on analyzing hotspot data as forest fire indicators are required for predicting hotspot occurrences. The objective of this work is to detect global and collective outliers on hotspot data in Riau Province in Sumatera Island for the period 2001-2012. The data used in this work are 4383 daily hotspots and 144 monthly hotspots. The method applied to discover outliers is the k-means clustering algorithm. The best clustering results are obtained on the number of clusters of 10 and the sum of squared error value is 18526.14. Based on the clustering results, we obtain 59 collective outliers and 30 global outliers on the hotspot dataset. The outliers on the hotspot data mostly occur in February, March, June, July, and August. The average frequency of outliers is 482.22 and the highest frequency of outliers is occurred in 2005. As many 1118 hotspots were found in the northern part of the Riau province on 21 June 2005. In August 2005 outliers spread on the whole area of Riau Province. For the period 2001-2012 there are no outliers occurred in April, November and December. This information is essential for an early warning system in forest fires prevention.

K-means clustering visualization of web-based OLAP operations for hotspot data

In the previous work we developed the web-based OLAP (On-line Analytical Processing) integrated with the data warehouse for hotspot data in Indonesia. This work aims to develop a visualization module for hotspot clusters resulted from OLAP operations including roll up and drill down. The data warehouse consists of hotspot data represented in multidimensional model with two dimensions: time and location. In the dimension time, the ordered sequence of elements from the higher-level of hierarchy to the lowest is from year, quarter, to month. Whereas, the sequence in the dimension location is from island, province, to district. The clustering algorithm we applied was K-means in which the best clustering was obtained for the size of cluster 4 with average value of SSE (sum of square error) 0.2944 for combinations of elements in the dimension time and location. Hotspot clusters are visualized in form of maps in addition to crosstabs and graphics built in the previous work. The map module in the web-based OLAP can be used to better organize and analyze the hotspot data as one of indicators for forest fires occurrence in Indonesia.

A TIME-DELAY CASCADING NEURAL NETWORK ARCHITECTURE FOR MODELING TIME-DEPENDENT PREDICTOR IN ONSET PREDICTION

The occurrence of rain before the real start of a r ainy season often mislead farmers into thinking tha t rainy season has started and suggesting them to start pla nting immediately. In reality, rainy season has not started yet, causing the already-planted rice seed to exper ience dehydration. Therefore, a model that can pred ict the onset of rainy season is required, so that draught disaster can be avoided. This study presents Time D elayCascading Neural Network (TD-CNN) which deals with situations where the response variable is determined by a number of time-dependent inter-rela ted predictors. The proposed model is used to predi ct the onset in Pacitan District Indonesia based on So uthern Oscillation Index (SOI). The Leave One Out (LOO) cross-validation with series data 1982-2012 a re used in order to compare the accuracy of the proposed model with the Back-Propagation Neural Network (BPNN) and Cascading Neural Network (CNN). The experiment shows that the accuracy of the proposed model is 0.74, slightly above than the t wo other models, BPNN and CNN which are 0.71 and 0.72, respectively.

Data warehouse and web-based OLAP for hotspot distribution in Indonesia

This work aims to develop a web-based OLAP (Online Analytical Processing) integrated with a data warehouse for hotspot distribution data. The data warehouse development adopts the three-tier data warehouse architecture. The data are represented in multidimensional model using the star scheme which consists of one data cube with two dimension tables i.e. the dimension time and the dimension location, and number of hotspots as the measure. The application applied OLAP operations including roll-up and drill-down and the results are represented in form of crosstabs and graphs. The concept hierarchy defines the sequence mappings in each dimension. In the dimension time, the sequence ordered from the higher-level of hierarchy to the lowest from year, quarter, to month. Whereas, the sequence in the dimension location is from island, province, to regency. The web-based OLAP can be used to better organize and analyze the hotspot data as well as provide a decision-making supporting in forest fire management.

Outlier Detection on Hotspots Data in Riau Province using DBSCAN Algorithm

Indonesia has serious problems in forest fires. One of the potential factors which indicates forest fires is hotspot. Hotspot is a forest fires indicator that detects a location with relatively higher temperature in comparison with nearby positions. One possible prevention efforts for forest fires is by detecting outliers on hotspots data. This study detects outlier on hotspots data in Riau Province in between year 2001 to 2012 using the DBSCAN algorithm and determines the distribution of outlier hotspots by region and time. The experiment results show that the highest occurrence of outliers is in 2005. The number of outliers on hotspots data reaches 1241 hotspots with the sum of square error (SSE) is 0.084. Outlier hotspots in Riau Province in 2005 spread across 11 districts/cities and 136 districts. In 2005 the highest outlier are found in Rokan Hulu with the number of outliers is 186 points. The highest frequency of hotspot that is considered as outliers is found in August 2005, with a total of 355 outliers in which as many 97 of these outliers are occurred in Rokan Hulu District.

Hotspot sequential pattern visualization in peatland of Sumatera and Kalimantan using shiny framework

Fires on peatland frequently occurred in Sumatra and Kalimantan. Fires on peatland can be identified by hotspot sequential patterns. Sequential pattern mining is one of data mining techniques that can be used to analyse hotspot sequential patterns. Sequential pattern discovery equivalent classes (SPADE) algorithm can be applied to extract hotspot sequential patterns. The objectives of this work are: 1) to obtain hotspot sequential pattern in Sumatra and Kalimantan in 2014 and 2015, and 2) to develop a web based application using Shiny framework that is available in R package for hotspot sequential pattern visualization in peatland of Sumatra and Kalimantan. Hotspot sequential patterns were obtained using minimum support of 0.01 with the focus of analysis is the hotspot sequences with length two or more events. This work generated as many 89 sequences with length 2 or more in Sumatra in 2014, 147 sequences in Sumatra in 2015, 48 sequences in Kalimantan in 2014, and 51 sequences in Kalimantan in 2015. Hotspot sequential patterns are visualized based on peatlands characteristics, weather, and social economy. The features in this web based application have been tested and the results show that all features work properly according to the test scenario.

Web-Based Application for Outliers Detection on Hotspot Data Using K-Means Algorithm and Shiny Framework

Outliers analysis on hotspot data as an indicator of fire occurences in Riau Province between 2001 and 2012 have been done, but it was less helpful in fire prevention efforts. This is because the results can only be used by certain people and can not be easily and quickly accessed by users. The purpose of this research is to create a web-based application to detect outliers on Hotspot data and to visualize the outliers based on the time and location. Outliers detection was done in the previous research using the k-means clustering method with global and collective outlier approach in Riau Province Hotspot data between 2001 and 2012. This work aims to develop a web-based application using the framework Shiny with the R programming language. This application provides several functions including summary and visualization of the selected data, clustering hotspot data using k-means algorithm, visualization of the clustering results and sum square error (SSE), and displaying global and collective outliers and visualization of outlier spread on Riau Province Map.