From 6b94b3b51cffa72493ed60638d3a0ca50c47b876 Mon Sep 17 00:00:00 2001
From: Dat Le <thedat.lenguyen@gmail.com>
Date: Mon, 23 Nov 2015 10:39:39 +0800
Subject: [PATCH] Added kaggle_geomean.py

Ref: https://github.com/MLWave/Kaggle-Ensemble-Guide/issues/7
https://en.wikipedia.org/wiki/Geometric_mean
---
 README.md                  | 16 +++++++++++++++-
 kaggle_geomean.py          | 31 +++++++++++++++++++++++++++++++
 samples/kaggle_geomean.csv |  6 ++++++
 3 files changed, 52 insertions(+), 1 deletion(-)
 create mode 100644 kaggle_geomean.py
 create mode 100644 samples/kaggle_geomean.csv

diff --git a/README.md b/README.md
index 73d43d2..70ca82e 100644
--- a/README.md
+++ b/README.md
@@ -31,6 +31,12 @@ For more information: http://mlwave.com/kaggle-ensembling-guide/
     parsing: ./samples/method3.csv
     wrote to ./samples/kaggle_avg.csv
 
+    $ python kaggle_geomean.py  "./samples/method*.csv" "./samples/kaggle_geomean.csv"
+    parsing: ./samples/method1.csv
+    parsing: ./samples/method2.csv
+    parsing: ./samples/method3.csv
+    wrote to ./samples/kaggle_geomean.csv
+
 ## Result:
 
     ==> ./samples/method1.csv <==
@@ -79,4 +85,12 @@ For more information: http://mlwave.com/kaggle-ensembling-guide/
     2,0
     3,9
     4,2
-    5,3
\ No newline at end of file
+    5,3
+
+    ==> ./samples/kaggle_geomean.csv <==
+    ImageId,Label
+    1,1.587401
+    2,0.000000
+    3,7.862224
+    4,3.301927
+    5,3.000000
diff --git a/kaggle_geomean.py b/kaggle_geomean.py
new file mode 100644
index 0000000..589bd18
--- /dev/null
+++ b/kaggle_geomean.py
@@ -0,0 +1,31 @@
+from __future__ import division
+from collections import defaultdict
+from glob import glob
+import sys
+import math
+
+glob_files = sys.argv[1]
+loc_outfile = sys.argv[2]
+
+def kaggle_bag(glob_files, loc_outfile, method="average", weights="uniform"):
+  if method == "average":
+    scores = defaultdict(float)
+  with open(loc_outfile,"wb") as outfile:
+    for i, glob_file in enumerate( glob(glob_files) ):
+      print "parsing:", glob_file
+      # sort glob_file by first column, ignoring the first line
+      lines = open(glob_file).readlines()
+      lines = [lines[0]] + sorted(lines[1:])
+      for e, line in enumerate( lines ):
+        if i == 0 and e == 0:
+          outfile.write(line)
+        if e > 0:
+          row = line.strip().split(",")
+          if scores[(e,row[0])] == 0:
+            scores[(e,row[0])] = 1
+          scores[(e,row[0])] *= float(row[1])
+    for j,k in sorted(scores):
+      outfile.write("%s,%f\n"%(k,math.pow(scores[(j,k)],1/(i+1))))
+    print("wrote to %s"%loc_outfile)
+
+kaggle_bag(glob_files, loc_outfile)
diff --git a/samples/kaggle_geomean.csv b/samples/kaggle_geomean.csv
new file mode 100644
index 0000000..fb6d7f3
--- /dev/null
+++ b/samples/kaggle_geomean.csv
@@ -0,0 +1,6 @@
+ImageId,Label
+1,1.587401
+2,0.000000
+3,7.862224
+4,3.301927
+5,3.000000