simonjisu / SMILE

Stock Movement predIction with Latent Embedding optimization

😊 SMILE - Stock Movement predIction with Latent Embedding optimization

People

• Yejin Hwang:
  • 🏫: SNU Graduate School of Data Science
  • 📧: [e-mail]
• Jisoo Jang:
  • 🏫: SNU Graduate School of Data Science
  • 📧: [e-mail]
• Minseok Chae:
  • 🏫: SNU Department of Mechanical Engineering
  • 📧: [e-mail]
• Taehun Kim:
  • 🏫: SNU Department of Mechanical Engineering
  • 📧: [e-mail]

Prerequisite

For pipenv user

# this command will automatically generate lock file by your system
$ pipenv lock
# install the env
$ pipenv install

For other users

# window users
pip install -r requirements_win.txt
# other users
pip install -r requirements.txt

How-to-Run

1. First write a experiment yaml file(see settings.yml in the experiments folder
   • There are also some pre-defined settings see the comments in the settings.yml
2. Run the experiments

   $ sh ./run_train.sh [your_exp_filename]
   # e.g.
   $ sh ./run_train.sh acl18.0

You can also run test on all the experiments by

$ sh ./run_test.sh [number of meta test for each iteration]
# e.g.
$ sh ./run_test.sh 100

Experiment Results

ACL18

KDD17

Table

Experiment File	Experiment	Test Type	Test Accuracy	Test Loss	Train Accuracy	Train Loss
acl18.0	acl18-baseline	test1	0.5118	7.4329	0.8633	0.3078
		test2	0.6877	3.8232
		test3	0.5135	6.6364
acl18.1	acl18-nohis	test1	0.5113	6.1526	0.8933	0.2813
		test2	0.6567	4.0848
		test3	0.5148	6.1174
acl18.2	acl18-n_finetune5	test1	0.5168	6.2921	0.8350	0.3608
		test2	0.6402	4.0439
		test3	0.4868	6.3577
acl18.3	acl18-n_layer1	test1	0.5040	6.0520	0.8850	0.2911
		test2	0.5545	4.9809
		test3	0.5008	5.7756
acl18.4	acl18-n_stock1.n_sample5	test1	0.5095	4.7025	0.8100	0.4778
		test2	0.5390	4.1183
		test3	0.5170	4.3116
acl18.5	acl18-n_stock1.n_sample10	test1	0.4810	5.6018	0.8475	0.3668
		test2	0.5423	4.4572
		test3	0.5100	5.0539
acl18.6	acl18-n_inner10	test1	0.5122	8.8750	0.8858	0.2751
		test2	0.7078	4.4255
		test3	0.5123	8.6040
kdd17.0	kdd17-baseline	test1	0.5127	3.1997	0.6550	0.6209
		test2	0.5958	2.6747
		test3	0.5188	3.2512
kdd17.1	kdd17-nohis	test1	0.5042	3.2292	0.6667	0.6026
		test2	0.5888	2.8033
		test3	0.5183	3.1743
kdd17.2	kdd17-n_finetune5	test1	0.5145	3.1745	0.6767	0.6149
		test2	0.5448	3.0430
		test3	0.5108	3.2819
kdd17.3	kdd17-n_layer1	test1	0.5063	3.4224	0.6650	0.6271
		test2	0.5532	3.3240
		test3	0.5110	3.3782
kdd17.4	kdd17-n_stock1.n_sample5	test1	0.4865	3.1318	0.6700	0.6482
		test2	0.5360	3.0204
		test3	0.4980	3.0609
kdd17.5	kdd17-n_stock1.n_sample10	test1	0.5110	3.1398	0.6500	0.6657
		test2	0.5475	3.0745
		test3	0.5145	3.1941
kdd17.6	kdd17-n_inner10	test1	0.5195	3.4049	0.6658	0.6209
		test2	0.6364	2.6134
		test3	0.5288	3.4379