Open AccessMultistage Tree Model for Crime Dataset in Iraq
Author(s) -
Reem Razzaq,
Abdulkadir Hussein,
Sadik Muayad,
Diaeldin A. Salih,
Mahdi Al-Qaraawi
Publication year - 2019
Publication title -
al-maǧallaẗ al-ʻirāqiyyaẗ li-handasaẗ al-ḥāsibāt wa-al-ittiṣālāt wa-al-sayṭaraẗ wa-al-naẓm
Language(s) - English
Resource type - Journals
eISSN - 2617-3352
pISSN - 1811-9212
DOI - 10.33103/uot.ijccce.19.2.1
Subject(s) - correlation , tree (set theory) , decision tree , rank correlation , copula (linguistics) , dependency (uml) , mathematics , statistics , decision tree model , computer science , correlation coefficient , tree structure , curse of dimensionality , data mining , algorithm , artificial intelligence , econometrics , binary tree , combinatorics , geometry
This research deals with the using of correlation measurement that leads to describing the degree of relationship between variables, quantities or qualities. Therefore, we implement a simple correlation coefficient and conditional correlation to introduce a regular vine copula, which gives different tree structures. Two methods to select tree structures are introduced. The first one adopts the Partial Correlation Constant (PCC) with constant, while the second method depends on the estimation of summation pathway.The proposed method makes modification on Diβmann’s algorithm to increase thedependency on each level of the tree using rank correlation measurement. Both methodsare adopted to construct the best model with more than three dimensions based on theavailable label crime dataset in Iraq. The selected model is used for selecting the suitabletree model and generating a decision with the low dimensionality of variables.