\relax \citation{boehm81} \citation{kemerer87} \citation{clcs03} \citation{clcs03} \citation{jorg04,jorgensen05} \citation{kemerer87,briand00,lum02,ferens98} \citation{witten05} \citation{Mendes2003} \citation{Mendes2003} \citation{baker07} \citation{Menzies2006} \citation{Li2009a} \citation{keung2008b} \citation{chang74} \citation{keung2008b} \citation{Li2009} \citation{hall03} \citation{gama06} \citation{FayIra93Multi} \citation{YanWeb02Comparative} \citation{Lipowezky1998} \citation{Walkerden1999} \citation{Kirsopp2002} \citation{Mendes2003} \citation{Li2009} \citation{baker07} \@writefile{toc}{\contentsline {section}{\numberline {I}Introduction}{1}} \citation{shepperd01} \citation{shepperd01} \citation{boehm81} \citation{boehm86} \@writefile{lof}{\contentsline {figure}{\numberline {1}{\ignorespaces There are many methods for effort estimation. One paper cannot hope to survey them all (e.g. this paper just explores 158). But to get a feel for the space of possibilities, consider the above list of design options for instance-based effort estimation. If we try all the following $N*W*D*M*R*K$ possibilities, this generates a space of $3*9*5*3*6*5>12,000$ methods. }}{2}} \newlabel{fig:cbr}{{1}{2}} \citation{me06d} \citation{park88} \citation{jensen83} \citation{putnam92} \citation{endres03} \@writefile{toc}{\contentsline {section}{\numberline {II}Background}{3}} \citation{gregor02} \citation{anderson85} \citation{simons99} \citation{shepperd07,jorgensen05} \citation{park88} \citation{jensen83} \citation{putnam92} \citation{chulani99} \citation{boehm81} \citation{boehm81} \citation{boehm81} \@writefile{toc}{\contentsline {section}{\numberline {III}Related Work}{5}} \@writefile{toc}{\contentsline {subsection}{\numberline {III-A}Regression-Based COCOMO}{5}} \newlabel{eq:coc1}{{1}{5}} \@writefile{lof}{\contentsline {figure}{\numberline {2}{\ignorespaces { The $f_i$ features used in this study. From\nobreakspace {}\cite {boehm81}. Most range from 1 to 6 representing ``very low'' to ``extremely high''. }}}{5}} \newlabel{fig:em}{{2}{5}} \citation{boehm81} \citation{boehm81} \citation{boehm81} \citation{boehm00b} \citation{boehm00b} \citation{shepperd07} \citation{li07} \@writefile{lof}{\contentsline {figure}{\numberline {3}{\ignorespaces The COCOMO-I $\beta _i$ table\nobreakspace {}\cite {boehm81}. For example, the bottom right cell is saying that if CPLX=6, then the nominal effort is multiplied by 1.65.}}{6}} \newlabel{fig:effortmults}{{3}{6}} \newlabel{eq:linear}{{2}{6}} \@writefile{toc}{\contentsline {subsection}{\numberline {III-B}Case-Based-Reasoning}{6}} \citation{kitch07} \citation{chulani99} \citation{shepperd97} \citation{miller02} \citation{kirsopp02} \citation{me05c} \citation{kleijnen97} \citation{ferens98} \citation{boehm00b} \citation{me04h} \citation{chang74} \citation{boehm00b} \citation{boehm81} \@writefile{toc}{\contentsline {section}{\numberline {IV}A Brief Tutorial on Row and Column Pruning}{7}} \citation{jalali07} \citation{baker07} \@writefile{lof}{\contentsline {figure}{\numberline {4}{\ignorespaces $NASA93$: intersection / union of examples in different data sets. }}{8}} \newlabel{fig:overlap1}{{4}{8}} \@writefile{toc}{\contentsline {section}{\numberline {V}Experiments}{8}} \@writefile{toc}{\contentsline {subsection}{\numberline {V-A}Data}{8}} \citation{kitch07} \citation{me06d} \citation{me06d} \@writefile{lof}{\contentsline {figure}{\numberline {5}{\ignorespaces $COC81$: intersection / union of examples in different data sets. }}{9}} \newlabel{fig:overlap2}{{5}{9}} \@writefile{toc}{\contentsline {subsection}{\numberline {V-B}Experimental Procedure}{9}} \citation{kitch07} \citation{mann47} \citation{kohavi97} \citation{quinlan92b} \citation{me06d} \citation{kohavi97} \citation{me06d} \@writefile{toc}{\contentsline {subsection}{\numberline {V-C}158 Methods}{10}} \citation{jalali07} \citation{me05c} \citation{boehm81,boehm00b} \citation{me05a} \citation{quinlan92b} \citation{me06d} \citation{shepperd97} \citation{myrtveit05} \@writefile{lof}{\contentsline {figure}{\numberline {6}{\ignorespaces Nine effort estimation methods explored in this paper. F is the number of features (columns) and P is the number of projects (rows).}}{11}} \newlabel{fig:abcd}{{6}{11}} \@writefile{toc}{\contentsline {subsection}{\numberline {V-D}Brief Notes on Nine Methods}{11}} \@writefile{lof}{\contentsline {figure}{\numberline {7}{\ignorespaces MRE results. Mann-Whitey (95\% confidence). These plots show number of losses of methods $(a,b,c,d,e,f,g,h,i)$ against 158 methods as judged by Mann-Whitney (95\% confidence). Each vertical set of marks shows results from 7 subsets of $COC81$ or 12 subsets of $NASA93$.}}{12}} \newlabel{fig:resultsBoth-MRE-Run1}{{7}{12}} \@writefile{lof}{\contentsline {figure}{\numberline {8}{\ignorespaces MER results. Mann-Whitey (95\% confidence). Same rig as Figure\nobreakspace {}7\hbox {}.}}{12}} \newlabel{fig:resultsBoth-MER-Run1}{{8}{12}} \@writefile{toc}{\contentsline {subsection}{\numberline {V-E}Methods Not Explored}{12}} \@writefile{lof}{\contentsline {figure}{\numberline {9}{\ignorespaces AR results, repeated three different times with three different random seeds. Same rig as Figure\nobreakspace {}7\hbox {}.}}{13}} \newlabel{fig:resultsBoth-AR-Run1}{{9}{13}} \@writefile{toc}{\contentsline {section}{\numberline {VI}Results}{13}} \citation{boehm81} \@writefile{toc}{\contentsline {section}{\numberline {VII}Discussion}{14}} \citation{shepperd97} \citation{jorg04} \citation{jorg04,jorgensen04,shepperd97,chulani99} \citation{boehm81} \citation{basili02} \@writefile{toc}{\contentsline {section}{\numberline {VIII}External Validity}{15}} \citation{chulani99} \citation{baker07} \citation{brieman96} \citation{FreSch97} \citation{me06d} \citation{boehm81} \@writefile{toc}{\contentsline {section}{\numberline {IX}Conclusion}{16}} \citation{boehm81} \citation{me06d} \@writefile{toc}{\contentsline {section}{Appendix}{17}} \@writefile{toc}{\contentsline {subsection}{\numberline {A}Data Used in This Study}{17}} \citation{me06d} \citation{quinlan92b} \citation{jalali07} \citation{kirsopp02} \citation{me05c} \citation{miller02} \citation{hall03} \citation{miller02} \citation{miller02} \citation{kohavi97} \@writefile{toc}{\contentsline {subsection}{\numberline {B}Learners Used in This Study}{18}} \@writefile{toc}{\contentsline {subsubsection}{\numberline {B.1}Learning with Model Trees}{18}} \@writefile{toc}{\contentsline {subsubsection}{\numberline {B.2}Other Learning Methods}{18}} \@writefile{toc}{\contentsline {subsection}{\numberline {C}Pre-Processors Used in This Study}{18}} \@writefile{toc}{\contentsline {subsubsection}{\numberline {C.1}Pre-processing with Row Pruning}{18}} \@writefile{toc}{\contentsline {subsubsection}{\numberline {C.2}Pre-Processing with Column Pruning}{18}} \citation{me06d,me05c} \citation{miller02} \citation{kohavi97} \citation{baker07} \citation{baker07} \bibstyle{IEEEtran} \bibdata{refs} \bibcite{boehm81}{1} \bibcite{kemerer87}{2} \bibcite{clcs03}{3} \bibcite{jorg04}{4} \bibcite{jorgensen05}{5} \bibcite{briand00}{6} \bibcite{lum02}{7} \bibcite{ferens98}{8} \bibcite{witten05}{9} \bibcite{Mendes2003}{10} \bibcite{baker07}{11} \bibcite{Menzies2006}{12} \bibcite{Li2009a}{13} \bibcite{keung2008b}{14} \bibcite{chang74}{15} \bibcite{Li2009}{16} \bibcite{hall03}{17} \bibcite{gama06}{18} \bibcite{FayIra93Multi}{19} \bibcite{YanWeb02Comparative}{20} \bibcite{Lipowezky1998}{21} \bibcite{Walkerden1999}{22} \bibcite{Kirsopp2002}{23} \bibcite{shepperd01}{24} \bibcite{boehm86}{25} \bibcite{me06d}{26} \bibcite{park88}{27} \bibcite{jensen83}{28} \bibcite{putnam92}{29} \bibcite{endres03}{30} \bibcite{gregor02}{31} \bibcite{anderson85}{32} \bibcite{simons99}{33} \bibcite{shepperd07}{34} \bibcite{chulani99}{35} \bibcite{boehm00b}{36} \bibcite{li07}{37} \@writefile{toc}{\contentsline {section}{References}{20}} \bibcite{kitch07}{38} \bibcite{shepperd97}{39} \bibcite{miller02}{40} \bibcite{kirsopp02}{41} \bibcite{me05c}{42} \bibcite{kleijnen97}{43} \bibcite{me04h}{44} \bibcite{jalali07}{45} \bibcite{mann47}{46} \bibcite{kohavi97}{47} \bibcite{quinlan92b}{48} \bibcite{me05a}{49} \bibcite{myrtveit05}{50} \bibcite{jorgensen04}{51} \bibcite{basili02}{52} \bibcite{brieman96}{53} \bibcite{FreSch97}{54}