Jurnal Valid

Poverty is one of the main problems faced by the people of Indonesia since ancient to the present so that a variety of planning, policy and program development has been and will be held on the point is to reduce the number of poor households. Family expectations, conditional cash assistance program, launched of Indonesia Government In 2007, first time in Indonesia. This Program aims to improve the quality of human beings by providing conditional cash assistance in accessing health and education services.The determination of the target family expectations, conditional cash assistance program done by BPS, for the first time used data of year 2005 owned ( based on the name and address ) subsequently in 2008 PPLS use 14 metric of poverty if household certain worthy relief. Virtue of analysis descriptive verification of data about households in the county of Jombang done by BAPEDA Jombang in 2010 that analysis needs desired or expected by majority of poverty in the county of Jombang expects direct aid, aid and assistance of venture capital, cattle the average who expects blt is worth 35,73 and as much as 28,16 % households in the county of Jombang hope for the help of livestock 17,99 hope for the help of venture capital, physical relief house at 8.28 %, aid daily needs 6.51 % and help or health education by 2,03 %.In this study of predictor variables have a lot then the MARS method can be used as tools that are expected to get the proper classification accuracy levels and to improve the accuracy of classification models and resampling methods used boosting where boosting is one method of ensemble used to improve the accuracy of classification by means of models generate a combination of a model get the best models and the variables that affect the best model with the method of binary response and MARS get accuracy classification needs help households with very poor and poor status of the method of binary response and boosting MARS. Households needs help is in the form of money and not in the form of money. Results of the research there were show by the method of MARS seven variables that affect household models are very poor ownership of residential buildings has the greatest influence on the model and the reduction of the value of the General Cross Validation (GCV), then a variable that has the most influence on models and substractiongcv to the poor is household income every month. Total accuracy classifications MARS and boosting MARS for households with the status of very poor is 70,50 % and 61,93 %, Results with total accuracy classifications MARS and boosting MARS to poor households with status is 64,30 % and 46,34 %.

Kata kunci :Multivariate Adaptive Regrression Spline, Boosting, Binary Response, Poverty

BAGIKAN