quant-stock-prediction

Quantitative Financeqlibrigorous codebase

Description

Quantitative Stock Prediction on CSI300

Objective

Design and implement a stock prediction model that forecasts next-day returns for CSI300 stocks. Your code goes in custom_model.py. Three reference implementations (LightGBM, LSTM, Transformer) are provided as read-only.

Evaluation

Signal quality: IC, ICIR, Rank IC. Portfolio (TopkDropout, top 50, drop 5): Annualized Return, Max Drawdown, Information Ratio. Evaluation is automatic via qlib's workflow.

Workflow Configuration

workflow_config.yaml lines 13-25 and 31-44 are editable. This is the model plus input-adapter/preprocessor block: you may change the dataset class (e.g., to TSDatasetH) or processors if your model needs a different input view. Instruments, date ranges, train/valid/test splits, and evaluation settings are fixed.

Code

custom_model.py

EditableRead-only

1# Custom stock prediction model for MLS-Bench
2#
3# EDITABLE section: CustomModel class with fit() and predict() methods.
4# FIXED sections: imports below.
5import numpy as np
6import pandas as pd
7import torch
8import torch.nn as nn
9import torch.nn.functional as F
10from qlib.model.base import Model
11from qlib.data.dataset import DatasetH
12from qlib.data.dataset.handler import DataHandlerLP
13
14DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
15

workflow_config.yaml

EditableRead-only

1# Qlib workflow configuration for CSI300 stock prediction benchmark.
2# Used by run_workflow.py — matches Alpha360/CSI300 official benchmark settings.
3
4qlib_init:
5  provider_uri: "~/.qlib/qlib_data/cn_data"
6  region: cn
7
8sys:
9  rel_path:
10    - "."           # So custom_model.py is importable via module_path
11
12task:
13  model:
14    class: CustomModel
15    module_path: custom_model

Additional context files (read-only):

qlib/qlib/model/base.py

Results

Model	Type	ic csi300 ↑	icir csi300 ↑	rank ic csi300 ↑	rank icir csi300 ↑	annualized return csi300 ↑	max drawdown csi300 ↓	information ratio csi300 ↑	ic csi100 ↑	icir csi100 ↑	rank ic csi100 ↑	rank icir csi100 ↑	annualized return csi100 ↑	max drawdown csi100 ↓	information ratio csi100 ↑	ic csi300 recent ↑	icir csi300 recent ↑	rank ic csi300 recent ↑	rank icir csi300 recent ↑	annualized return csi300 recent ↑	max drawdown csi300 recent ↓	information ratio csi300 recent ↑
lgbm	baseline	-	-	-	-	-	-	-	0.036	0.220	0.044	0.277	-0.017	-0.158	-0.334	0.025	0.182	0.041	0.302	0.020	-0.076	0.266
lgbm	baseline	0.040	0.308	0.049	0.402	0.020	-0.100	0.280	0.036	0.220	0.044	0.277	-0.017	-0.158	-0.334	0.025	0.182	0.041	0.302	0.020	-0.076	0.266
lgbm	baseline	0.041	0.307	0.050	0.403	0.013	-0.116	0.188	0.037	0.220	0.046	0.278	-0.015	-0.164	-0.269	0.024	0.178	0.041	0.301	0.056	-0.062	0.781
lgbm	baseline	0.040	0.308	0.049	0.402	0.020	-0.100	0.280	0.036	0.220	0.044	0.277	-0.017	-0.158	-0.334	0.025	0.182	0.041	0.302	0.020	-0.076	0.266
lgbm	baseline	0.040	0.308	0.049	0.402	0.020	-0.100	0.280	0.036	0.220	0.044	0.277	-0.017	-0.158	-0.334	0.025	0.182	0.041	0.302	0.020	-0.076	0.266
lstm	baseline	0.047	0.370	0.058	0.467	0.085	-0.104	1.031	0.039	0.213	0.044	0.246	-0.016	-0.176	-0.258	0.031	0.238	0.043	0.313	0.070	-0.069	0.885
transformer	baseline	0.012	0.074	0.035	0.245	-0.036	-0.166	-0.440	0.020	0.109	0.034	0.191	-0.059	-0.282	-1.012	0.001	0.010	0.027	0.183	-0.085	-0.160	-1.091
anthropic/claude-opus-4.6	vanilla	0.048	0.360	0.059	0.464	0.073	-0.093	1.213	0.032	0.162	0.044	0.233	-0.016	-0.199	-0.259	0.031	0.242	0.048	0.363	0.087	-0.032	1.247
google/gemini-3.1-pro-preview	vanilla	0.046	0.336	0.057	0.441	0.083	-0.058	1.267	0.034	0.182	0.046	0.261	-0.002	-0.198	-0.028	0.028	0.208	0.042	0.319	0.045	-0.046	0.670
gpt-5.4-pro	vanilla	0.007	0.035	0.026	0.132	-0.055	-0.357	-0.491	0.006	0.026	0.018	0.080	-0.096	-0.340	-1.411	0.007	0.049	0.034	0.231	-0.058	-0.144	-0.788
anthropic/claude-opus-4.6	agent	0.048	0.360	0.059	0.464	0.073	-0.093	1.213	0.032	0.162	0.044	0.233	-0.016	-0.199	-0.259	0.031	0.242	0.048	0.363	0.087	-0.032	1.247
google/gemini-3.1-pro-preview	agent	0.049	0.337	0.064	0.467	0.083	-0.110	1.057	0.048	0.258	0.059	0.338	-0.001	-0.115	-0.012	0.029	0.225	0.048	0.351	0.098	-0.050	1.435
gpt-5.4-pro	agent	0.040	0.311	0.051	0.411	0.032	-0.098	0.464	0.033	0.195	0.043	0.256	-0.028	-0.180	-0.518	0.023	0.161	0.041	0.291	0.017	-0.071	0.224

Agent Conversations

anthropic/claude-opus-4.6

5 steps

google/gemini-3.1-pro-preview

6 steps

gpt-5.4-pro

6 steps