From 81c1d63dc1d81057e22322caf519fc8bb99bb58d Mon Sep 17 00:00:00 2001
From: Jon Alberdi <j.alberdi@criteo.com>
Date: Thu, 7 Mar 2019 15:12:39 +0100
Subject: [PATCH] Use seed in train_test_split

Permits having the same result when running
the same experiment twice.

Follows recommendation given at
https://github.com/coreylynch/pyFM/issues/19
---
 pyfm/pylibfm.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/pyfm/pylibfm.py b/pyfm/pylibfm.py
index b0b71a4..5926106 100644
--- a/pyfm/pylibfm.py
+++ b/pyfm/pylibfm.py
@@ -174,7 +174,7 @@ def fit(self, X, y):
         if self.verbose == True:
             print("Creating validation dataset of %.2f of training for adaptive regularization" % self.validation_size)
         X_train, validation, train_labels, validation_labels = train_test_split(
-            X, y, test_size=self.validation_size)
+            X, y, test_size=self.validation_size, random_state=self.seed)
         self.num_attribute = X_train.shape[1]
 
         # Convert datasets to sklearn sequential datasets for fast traversal