Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Topic/add readme #2

Open
wants to merge 4 commits into
base: master
Choose a base branch
from
Open

Topic/add readme #2

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
f907e87
Update README and add additional script for random forest model
Mar 18, 2019
dcf1143
updating rf_model.r
Mar 18, 2019
2a38f19
Updating README
Mar 18, 2019
99e9f24
Merge branch 'master' into topic/add_readme
AmreenS Mar 18, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 13 additions & 2 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,8 @@ crossRaterRelibility.clean.txt - Data provided to raters to create labels. Data

### Code

rf_model.r - Script to build random forest models, predict and cross-validate.
Experiments to understand effects of augmenting fingerprint metrics to string similarity.

rf.r - Various debugging/exploratory code used to build random forest models, predict and
cross-validate. Also icludes experiments to understand effects of augmenting fingerprint
Expand All @@ -17,8 +19,17 @@ rf.r - Various debugging/exploratory code used to build random forest models, pr

### Models

#To be uploaded
rfmodelsFullP7.c.RData - random forest model

Models can be found at https://drive.google.com/drive/folders/170vwBToJDdBZ6li_ag2vKeqxr2ni03eM?usp=sharing

match2345_pairs.RDS - all pairs feature dataframe

rfmodels.Rdata -

rfmodelsC.Rdata -

rfmodelsFullP7.c.RData -

HOMONYM_RFMODELS - Contains all random forest models built using fingerprint (files, projects, text) metrics used for cross validation.


Expand Down
168 changes: 168 additions & 0 deletions rf_model.r
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,168 @@
tos all
#try random forest
library(randomForest);

pairs <- readRDS("../MODELS_PHASE5/match2345_pairs.RDS");

pairs$pairs$im = as.factor(pairs$pairs$"is_match");
mm = pairs$pairs$"is_match" == 1;
nmm=!mm & apply(pairs$pairs[,c("n","e","ln","fn","un","ifn","ln1","fn1")], 1, max) > .8 | pairs$pairs$ad>0|pairs$pairs$tdz>.9|pairs$pairs$d2vSim>0;
nnmm=!mm & !nmm
vv = 1:dim(pairs$pairs)[1];


#Fit RF Model with fingerprints

rf = list();
p = list();
sel=sample(0:9,sum(mm),replace=T);
sel1=sample(0:9,sum(nmm),replace=T);
sel2=sample(0:9,sum(nnmm),replace=T);

for (i in 0:9){
mm1t = vv[mm][sel!=i];
mm1v = vv[mm][sel==i];
mm0t = vv[nmm][sel1!=i];
mm0v = vv[nmm][sel1==i];
mm0v = c(mm0v, vv[nnmm][sel2==i])
dt=pairs$pairs[c(mm1t,mm0t),c("n","e","ln","fn","un","d2vSim", "ad", "tdz", "ifn", "ln1f", "fnf", "ln1","fn1", "im")];
dv=pairs$pairs[c(mm1v,mm0v),c("n","e","ln","fn","un","d2vSim","ad", "tdz", "ifn", "ln1f", "fnf", "ln1","fn1","im")];
rf[[i+1]] = randomForest(im ~ ., dt, importance=T);
p[[i+1]] = predict(rf[[i+1]],dv);
print (table(p[[i+1]],dv$im));
}

#inspect errors
res = c();
for (i in 0:9){
mm1v = vv[mm][sel==i];
mm0v = vv[nmm][sel1==i];
mm0v = c(mm0v, vv[nnmm][sel2==i])
dv=pairs$pairs[c(mm1v,mm0v),c("n","e","ln","fn","un","d2vSim","ad", "tdz", "ifn", "ln1f", "fnf", "ln1","fn1","im","id1","id2")];
miss = dv[dv$im!=p[[i+1]],c("id1","id2","im")];
ss = match(paste(pairs$pairs$id1, pairs$pairs$id2,sep=";"),paste(miss$id1,miss$id2,sep=";"),nomatch=0)>0;
misd = pairs$pairs[ss,];
aa = paste(pairs$data1[misd$id1,"a"],pairs$data2[misd$id2,"a"], sep=";");
res = rbind(res, cbind(misd[,c("id1","id2","d2vSim","ad", "tdz","im")], aa));
}
#save
save(rf, p, res, sel, sel1, sel2, file = "/home/audris/rfmodels.RData");

res[res$im==0,c("id1","id2","d2vSim","ad","tdz","im")];

#manually correct errors
eq= c(1817022,2620775,3198135,3823436,4923380,2469101,5117770,429508 ,1674130,2018882,2469100,
2863731,4667324,5063908,5351827,1674131 ,2566524,3924916,4669669,5028394,5066253,
1814677,3198134,3305223,4454519,4603843,3647564,3943395,4452175,4667325,1425379,2552916,
4452174,5026049,5030739,5115306,1814676,2021226,2550570);

pairs$pairs$im=as.factor(pairs$pairs$"is_match");
fr=c(1117,1771,2183,1632,2265,2265,1221, 666,1900,487,1053,1075,1085,1085,1207,1411,748, 1411, 1221, 1632, 1076, 1148, 748, 774, 775, 774, 1115, 1053, 981, 185, 1093, 1899, 2099, 373, 966, 1094, 184, 2100, 2144, 424, 425, 1450, 1228, 1116, 775, 1117, 1364, 774, 1086, 1900, 1118, 1227, 1227);
to=c(1411, 666, 980,1085, 749, 748,487,1772,1365,1221,2160,1207,1632,1631,1076,1117,2264,1118, 486, 1086, 1207, 528, 2264, 1992, 1992, 775, 1409, 2161, 2183, 373, 1096, 1365, 1225, 184, 1619,1096, 373, 1225, 2255, 229, 230, 1683, 2099, 1410, 1991, 1410, 1899, 1991, 1631, 1364, 1410, 2100, 2099);
for (i in c(1:length(fr))){
pairs$pairs$im[pairs$pairs$id1==fr[i]&pairs$pairs$id2==to[i]]=1;
pairs$pairs$im[pairs$pairs$id2==fr[i]&pairs$pairs$id1==to[i]]=1;
}
pp=pairs$pairs[eq,]
pp=pp[pp$im==0,]
paste(pairs$data1[pp$id1,"a"],pairs$data1[pp$id2,"a"])

for (i in c(1:length(pp$id1))){
pairs$pairs$im[pairs$pairs$id1==pp$id1[i]&pairs$pairs$id2==pp$id2[i]]=1;
pairs$pairs$im[pairs$pairs$id2==pp$id1[i]&pairs$pairs$id1==pp$id2[i]]=1;
}

pp=pairs$pairs[pairs$pairs$im!=pairs$pairs$"is_match",];
pp=pp[pp$im==1&pp$id1<pp$id2,];
paste(pairs$data1[pp$id1,"a"],pairs$data1[pp$id2,"a"]);


#Fit again after correcting errors
mmC = pairs$pairs$im == 1;
nmmC=!mmC & apply(pairs$pairs[,c("n","e","ln","fn","un","ifn","ln1","fn1")], 1, max) > .8 | pairs$pairs$ad>0|pairs$pairs$tdz>.9|pairs$pairs$d2vSim>0;
nnmmC=!mmC & !nmmC
selC=sample(0:9,sum(mmC),replace=T);
selC1=sample(0:9,sum(nmmC),replace=T);
selC2=sample(0:9,sum(nnmmC),replace=T);
rfC = list();
pC = list();
for (i in 0:9){
mm1t = vv[mmC][selC!=i];
mm1v = vv[mmC][selC==i];
mm0t = vv[nmmC][selC1!=i];
#mm0t = c(mm0t, vv[nnmmC][selC2!=i]) #mm0t = c(mm0t, vv[nnmmC&fullP!=pairs$pairs$im])
mm0v = vv[nmmC][selC1==i];
mm0v = c(mm0v, vv[nnmmC][selC2==i])#include the entire validation set for prediction
dt=pairs$pairs[c(mm1t,mm0t),c("n","e","ln","fn","un","d2vSim", "ad", "tdz", "ifn", "ln1f", "fnf", "ln1","fn1", "im")];
dv=pairs$pairs[c(mm1v,mm0v),c("n","e","ln","fn","un","d2vSim","ad", "tdz", "ifn", "ln1f", "fnf", "ln1","fn1","im")];
rfC[[i+1]] = randomForest(im ~ ., dt, importance=T);
pC[[i+1]] = predict(rfC[[i+1]],dv);
print (table(pC[[i+1]],dv$im));
}

resC = c();
for (i in 0:9){
mm1v = vv[mmC][selC==i];
mm0v = vv[nmmC][selC1==i];
mm0v = c(mm0v, vv[nnmmC][selC2==i])#include the entire validation set for prediction
dv=pairs$pairs[c(mm1v,mm0v),c("n","e","ln","fn","un","d2vSim","ad", "tdz", "ifn", "ln1f", "fnf", "ln1","fn1","im","id1","id2")];
miss = dv[dv$im!=pC[[i+1]],c("id1","id2","im")];
ss = match(paste(pairs$pairs$id1, pairs$pairs$id2,sep=";"),paste(miss$id1,miss$id2,sep=";"),nomatch=0)>0;
misd = pairs$pairs[ss,];
aa = paste(pairs$data1[misd$id1,"a"],pairs$data2[misd$id2,"a"], sep=";");
resC = rbind(resC, cbind(misd[,c("id1","id2","d2vSim","ad", "tdz","im")], aa));
}
#save
save(rfC, pC, resC, selC, selC1, selC2, file = "/home/audris/rfmodelsC.RData");
load(file = "/home/audris/rfmodelsC.RData");

resC[resC$im==0,];


#try simpler models that can be applied on the entire universe
rfCS = list();
pCS = list();
for (i in 0:9){
mm1t = vv[mmC][selC!=i];
mm1v = vv[mmC][selC==i];
mm0t = vv[nmmC][selC1!=i];
#mm0t = c(mm0t, vv[nnmmC][selC2!=i]) #mm0t = c(mm0t, vv[nnmmC&fullP!=pairs$pairs$im])
mm0v = vv[nmmC][selC1==i];
mm0v = c(mm0v, vv[nnmmC][selC2==i])#include the entire validation set for prediction
dt=pairs$pairs[c(mm1t,mm0t),c("n","e","ln","fn","un", "ifn", "d2vSim", "im")];
dv=pairs$pairs[c(mm1v,mm0v),c("n","e","ln","fn","un", "ifn", "d2vSim", "im")];
rfCS[[i+1]] = randomForest(im ~ ., dt, importance=T);
pCS[[i+1]] = predict(rfCS[[i+1]],dv);
print (table(pCS[[i+1]],dv$im));
}

resCS = c();
for (i in 0:9){
mm1v = vv[mmC][selC==i];
mm0v = vv[nmmC][selC1==i];
mm0v = c(mm0v, vv[nnmmC][selC2==i])#include the entire validation set for prediction
dv=pairs$pairs[c(mm1v,mm0v),c("n","e","ln","fn","un","d2vSim","ad", "tdz", "ifn", "ln1f", "fnf", "ln1","fn1","im","id1","id2")];
miss = dv[dv$im!=pCS[[i+1]],c("id1","id2","im")];
ss = match(paste(pairs$pairs$id1, pairs$pairs$id2,sep=";"),paste(miss$id1,miss$id2,sep=";"),nomatch=0)>0;
misd = pairs$pairs[ss,];
aa = paste(pairs$data1[misd$id1,"a"],pairs$data2[misd$id2,"a"], sep=";");
resCS = rbind(resCS, cbind(misd[,c("id1","id2","d2vSim","im")], aa));
}
save(rfCS, pCS, resCS, pairs, file = "/home/audris/rfmodelsCS.RData");

#try even simpler model without the fingerprints
rfCSB = list();
pCSB = list();
for (i in 0:9){
mm1t = vv[mmC][selC!=i];
mm1v = vv[mmC][selC==i];
mm0t = vv[nmmC][selC1!=i];
#mm0t = c(mm0t, vv[nnmmC][selC2!=i]) #mm0t = c(mm0t, vv[nnmmC&fullP!=pairs$pairs$im])
mm0v = vv[nmmC][selC1==i];
mm0v = c(mm0v, vv[nnmmC][selC2==i])#include the entire validation set for prediction
dt=pairs$pairs[c(mm1t,mm0t),c("n","e","ln","fn","un", "ifn", "im")];
dv=pairs$pairs[c(mm1v,mm0v),c("n","e","ln","fn","un", "ifn", "im")];
rfCSB[[i+1]] = randomForest(im ~ ., dt, importance=T);
pCSB[[i+1]] = predict(rfCSB[[i+1]],dv);
print (table(pCSB[[i+1]],dv$im));
}