Extract raw files from the Temporal-Relational-Ranking-for-Stock-Prediction repository.
cd dataset
tar zxvf raw.tar.gz
Pre-process and clean raw files.
cd dataset
python process.py
python train.py --nworkers 2 --use-amp --market NYSE --sequence-length 8 --add-self-loop --relational-graph wiki --model SimpleFinSIR --nhidden 8 --recurrent-layers 1 --recurrent-dropout 0 --relational-agg sym --relational-dropout 0 --readout-layers 1 --readout-dropout 0 --epochs 50 --lr 1e-3 --wd 1e-5 --factor 0.5 --patience 50 --alpha 0.1
python train.py --nworkers 2 --use-amp --market NYSE --sequence-length 8 --add-self-loop --relational-graph wiki --model FinSIR --nhidden 16 --recurrent-layers 1 --recurrent-dropout 0 --relational-agg sym --relational-dropout 0 --readout-layers 1 --readout-dropout 0 --epochs 50 --lr 1e-3 --wd 1e-5 --factor 0.5 --patience 50 --alpha 0.1
python train.py --nworkers 2 --use-amp --market NYSE --sequence-length 16 --add-self-loop --relational-graph industry --model SimpleFinSIR --nhidden 64 --recurrent-layers 1 --recurrent-dropout 0 --relational-agg sym --relational-dropout 0 --readout-layers 1 --readout-dropout 0 --epochs 50 --lr 1e-3 --wd 1e-5 --factor 0.5 --patience 50 --alpha 0.1
python train.py --nworkers 2 --use-amp --market NYSE --sequence-length 8 --add-self-loop --relational-graph industry --model FinSIR --nhidden 8 --recurrent-layers 1 --recurrent-dropout 0 --relational-agg sym --relational-dropout 0 --readout-layers 1 --readout-dropout 0 --epochs 50 --lr 1e-3 --wd 1e-5 --factor 0.5 --patience 50 --alpha 0.1
python train.py --nworkers 2 --use-amp --market NASDAQ --sequence-length 2 --add-self-loop --relational-graph wiki --model SimpleFinSIR --nhidden 8 --recurrent-layers 1 --recurrent-dropout 0 --relational-agg sym --relational-dropout 0 --readout-layers 1 --readout-dropout 0 --epochs 50 --lr 1e-3 --wd 1e-5 --factor 0.5 --patience 50 --alpha 0.1
python train.py --nworkers 2 --use-amp --market NASDAQ --sequence-length 2 --add-self-loop --relational-graph wiki --model FinSIR --nhidden 32 --recurrent-layers 1 --recurrent-dropout 0 --relational-agg sym --relational-dropout 0 --readout-layers 1 --readout-dropout 0 --epochs 50 --lr 1e-3 --wd 1e-5 --factor 0.5 --patience 50 --alpha 1.0
python train.py --nworkers 2 --use-amp --market NASDAQ --sequence-length 2 --add-self-loop --relational-graph industry --model SimpleFinSIR --nhidden 16 --recurrent-layers 1 --recurrent-dropout 0 --relational-agg sym --relational-dropout 0 --readout-layers 1 --readout-dropout 0 --epochs 50 --lr 1e-3 --wd 1e-5 --factor 0.5 --patience 50 --alpha 0.1
python train.py --nworkers 2 --use-amp --market NASDAQ --sequence-length 2 --add-self-loop --relational-graph industry --model FinSIR --nhidden 16 --recurrent-layers 1 --recurrent-dropout 0 --relational-agg sym --relational-dropout 0 --readout-layers 1 --readout-dropout 0 --epochs 50 --lr 1e-3 --wd 1e-5 --factor 0.5 --patience 50 --alpha 0.1
| Model | IRR 1 | IRR 5 | MRR 1 | MRR 5 | MSE |
|---|---|---|---|---|---|
| RankLSTM | 0.0140 ± 0.2322 | 0.0605 ± 0.1096 | 0.0260 ± 0.0080 | 0.0168 ± 0.0040 | 2.2700e-04 ± 1.0000e-06 |
| RSR-I (Wiki) | 0.6148 ± 0.6462 | 0.4465 ± 0.2170 | 0.0265 ± 0.0066 | 0.0234 ± 0.0021 | 2.2700e-04 ± 1.0000e-06 |
| RSR-E (Wiki) | 0.9491 ± 0.4731 | 0.4075 ± 0.2395 | 0.0339 ± 0.0058 | 0.0226 ± 0.0015 | 2.2800e-04 ± 1.0000e-06 |
| SimpleFinSIR (Wiki) | 1.3457 ± 0.4256 | 0.5673 ± 0.1267 | 0.0369 ± 0.0068 | 0.0244 ± 0.0022 | 2.2800e-04 ± 1.0000e-06 |
| FinSIR (Wiki) | 1.6034 ± 0.3451 | 0.4337 ± 0.1251 | 0.0298 ± 0.0043 | 0.0200 ± 0.0027 | 2.2700e-04 ± 1.0000e-06 |
| RSR-I (industry) | 1.1937 ± 0.4384 | 0.4734 ± 0.1135 | 0.0348 ± 0.0062 | 0.0229 ± 0.0014 | 2.2700e-04 ± 1.0000e-06 |
| RSR-E (industry) | 1.2093 ± 0.3199 | 0.4335 ± 0.1344 | 0.0362 ± 0.0059 | 0.0235 ± 0.0020 | 2.2700e-04 ± 1.0000e-06 |
| SimpleFinSIR (industry) | 1.2739 ± 0.3823 | 0.4796 ± 0.1325 | 0.0343 ± 0.0045 | 0.0221 ± 0.0019 | 2.2800e-04 ± 1.0000e-06 |
| FinSIR (industry) | 1.4761 ± 0.4458 | 0.5338 ± 0.1504 | 0.0353 ± 0.0059 | 0.0246 ± 0.0022 | 2.2800e-04 ± 1.0000e-06 |
| Model | IRR 1 | IRR 5 | MRR 1 | MRR 5 | MSE |
|---|---|---|---|---|---|
| RankLSTM | 0.2882 ± 0.2457 | 0.1485 ± 0.1269 | 0.0340 ± 0.0068 | 0.0201 ± 0.0042 | 3.7800e-04 ± 1.0000e-06 |
| RSR-I (Wiki) | 0.2476 ± 0.3306 | 0.0630 ± 0.1417 | 0.0308 ± 0.0070 | 0.0175 ± 0.0038 | 3.7900e-04 ± 3.0000e-06 |
| RSR-E (Wiki) | 0.2085 ± 0.4255 | 0.2044 ± 0.0776 | 0.0299 ± 0.0079 | 0.0188 ± 0.0042 | 3.7900e-04 ± 1.0000e-06 |
| SimpleFinSIR (Wiki) | 1.1161 ± 0.3207 | 0.4460 ± 0.1685 | 0.0472 ± 0.0039 | 0.0262 ± 0.0042 | 3.7700e-04 ± 0.0000e+00 |
| FinSIR (Wiki) | 0.7838 ± 0.5050 | 0.3051 ± 0.1583 | 0.0408 ± 0.0050 | 0.0241 ± 0.0034 | 3.9500e-04 ± 3.5000e-05 |
| RSR-I (industry) | 0.5934 ± 0.3493 | 0.2983 ± 0.1274 | 0.0309 ± 0.0033 | 0.0220 ± 0.0020 | 3.8000e-04 ± 1.0000e-06 |
| RSR-E (industry) | 1.1114 ± 0.2703 | 0.5670 ± 0.0880 | 0.0462 ± 0.0052 | 0.0273 ± 0.0039 | 3.7800e-04 ± 1.0000e-06 |
| SimpleFinSIR (industry) | 0.9334 ± 0.3608 | 0.3106 ± 0.1055 | 0.0429 ± 0.0058 | 0.0242 ± 0.0046 | 3.7700e-04 ± 1.0000e-06 |
| FinSIR (industry) | 1.2307 ± 0.3761 | 0.6747 ± 0.1374 | 0.0487 ± 0.0050 | 0.0310 ± 0.0021 | 3.7800e-04 ± 1.0000e-06 |