diff --git a/03_predict.py b/03_predict.py
index 4662b9f..a58e69a 100755
--- a/03_predict.py
+++ b/03_predict.py
@@ -9,7 +9,6 @@
 import pandas as pd
 import numpy as np
 from sklearn.model_selection import train_test_split
-from sklearn.metrics import confusion_matrix, accuracy_score
 
 def main():
     parser = argparse.ArgumentParser()
@@ -21,8 +20,8 @@ def main():
 
     if not os.path.exists(args.modelpath):
         parser.error("ember model {} does not exist".format(args.modelpath))
-    if not os.path.exists(args.modelpath):
-        parser.error("ember model {} does not exist".format(args.output))
+    if not os.path.exists(args.output):
+        os.mkdir(args.output)
     if not os.path.exists(args.csv):
         parser.error("ember model {} does not exist".format(args.csv))
     if not os.path.exists(args.datadir):
@@ -55,7 +54,6 @@ def main():
                 y_pred.append(0)
                 errorcount += 1
                 
-
     #print and save accuracy
     y_pred_01 = np.array(y_pred)
     y_pred_01 = np.where(y_pred_01 > 0.75, 1, 0)   
@@ -72,10 +70,14 @@ def main():
 
     #save csv
     raw_predict = pd.DataFrame({'hash': _name, 'y': y, 'ypred': y_pred_01})
-    raw_predict.to_csv(os.path.join(args.output, 'predict_with_label.csv'), index=False)
+    raw_predict.to_csv(os.path.join(args.output, 'predict_with_label.csv'), index=False, header=None)
 
     r = pd.DataFrame({'hash': _name, 'y_pred': y_pred_01})
-    r.to_csv(os.path.join(args.output, 'result.csv'), index=False)
+    r.to_csv(os.path.join(args.output, 'result.csv'), index=False, header=None)
+
+    #print errorcount
+    print("Error : %d" % (errorcount))
     
 if __name__ == "__main__":
-    main()
\ No newline at end of file
+    main()
+    print("Done")
\ No newline at end of file
diff --git a/04_get_accuarcy.py b/04_get_accuarcy.py
index baf1858..e2e1868 100644
--- a/04_get_accuarcy.py
+++ b/04_get_accuarcy.py
@@ -5,15 +5,29 @@
 
 parser = argparse.ArgumentParser()
 parser.add_argument('-c', '--csv', type=str, required=True, help='csv file for getting accuracy')
+parser.add_argument('-t', '--threshold', type=str, default=0.75, help='threadshold for predicting')
 args = parser.parse_args()
 
 def main():
-    data = pd.read_csv(args.csv)
-
+    data = pd.read_csv(args.csv, names=['hash', 'y', 'ypred'])
+    
     y = data.y
-    ypred = np.where(np.array(data.ypred) > 0.75, 1, 0)
+    ypred = np.where(np.array(data.ypred) > args.threshold, 1, 0)
+
+    #get and print accuracy
     accuracy = accuracy_score(y, ypred)
-    print("accuracy : %.2f%%" % (np.round(accuracy, decimals=2)*100))
+    print("accuracy : %.0f%%" % (np.round(accuracy, decimals=2)*100))
+
+    #get and print matrix
+    mt = confusion_matrix(y, ypred)
+    t = mt[0][0]
+    mt[0][0] = mt[1][1]
+    mt[1][1] = t
+    print(mt)
+
+    #print FP, FN
+    print("False Postive : %.0f%%" % (round(mt[0][1]/(mt[0][1]+mt[1][1]), 2)*100))
+    print("False Negative : %.0f%%" % (round(mt[1][0]/(mt[0][0]+mt[1][0]), 2)*100))
 
 if __name__=='__main__':
     main()
\ No newline at end of file
diff --git a/README.ME b/README.ME
deleted file mode 100644
index 78afce6..0000000
--- a/README.ME
+++ /dev/null
@@ -1,32 +0,0 @@
-# Reference
-https://github.com/endgameinc/ember  
-
-H. Anderson and P. Roth, "EMBER: An Open Dataset for Training Static PE Malware Machine Learning Models”, in ArXiv e-prints. Apr. 2018.  
-
-```
-@ARTICLE{2018arXiv180404637A,  
-  author = {{Anderson}, H.~S. and {Roth}, P.},  
-  title = "{EMBER: An Open Dataset for Training Static PE Malware Machine Learning Models}",  
-  journal = {ArXiv e-prints},  
-  archivePrefix = "arXiv",  
-  eprint = {1804.04637},  
-  primaryClass = "cs.CR",  
-  keywords = {Computer Science - Cryptography and Security},  
-  year = 2018,  
-  month = apr,  
-  adsurl = {http://adsabs.harvard.edu/abs/2018arXiv180404637A},  
-}  
-```  
-
-# install
-above python 3.5    
-```
-#install virtualenv
-$ virtualenv emberenv -p python3
-$ . ./emberenv/bin/activate
-```
-  
-```
-(emberenv)$ pip instal -r requirements.txt
-(emberenv)$ pip install -U scikit-learn
-```
\ No newline at end of file
diff --git a/utils/upx_packer.py b/utils/upx_packer.py
new file mode 100755
index 0000000..9dd6278
--- /dev/null
+++ b/utils/upx_packer.py
@@ -0,0 +1,32 @@
+#require sudo apt install upx
+import os
+import argparse
+import subprocess
+import tqdm
+
+parser = argparse.ArgumentParser()
+parser.add_argument("-d", "--datadir", required=True, help="directory for packing")
+parser.add_argument("-o", "--output", required=True, help="output directory")
+args = parser.parse_args()
+
+if not os.path.exists(args.datadir):
+    parser.error("ember model {} does not exist".format(args.modelpath))
+if not os.path.exists(args.output):
+    os.makedirs(args.output)
+
+for _file in tqdm.tqdm(os.listdir(args.datadir)):
+    path = os.path.join(args.datadir, _file)
+    output = os.path.join(args.output, _file)
+
+    command = ['upx -l ' + path]
+    r = subprocess.Popen(command, stdout=subprocess.PIPE, shell=True, stderr=subprocess.STDOUT)
+    NotPacker = r.communicate()[0].decode('utf-8')
+    
+    if "NotPackedException" in  NotPacker:
+        command = ['upx -o ' + output + ' ' + path]
+        subprocess.Popen(command, stdout=subprocess.PIPE, shell=True, stderr=subprocess.STDOUT)
+    else:
+        command = ['cp ' + path + ' ' + output]
+        subprocess.Popen(command, stdout=subprocess.PIPE, shell=True, stderr=subprocess.STDOUT)
+    
+print("Done")