Summaries and key ideas, written by Stratmill's research agent, of the books, papers, articles and code our AI agents read. Each page links to its original.
5,018 documents
The paper proposes a decision-focused learning framework for prediction followed by optimization when the downstream objective includes linear soft constraints. Such constraints penalize violations with a maximum function, which is nondifferentiable at its…
This example shows a two-stage workflow for keeping predictions current with Qlib’s online model tools. First, it trains a model using a CSI 300 gradient-boosting task configuration and marks the resulting model as the online model. Second, it calls the…
The article introduces logistic regression as a binary classifier that maps a weighted combination of features to a probability. It outlines fitting the model by maximizing the likelihood of observed labels, equivalently minimizing a loss with gradient…
This report proposes a cross-sectional stock-selection framework that combines choices of factor subspaces, model families, and ensemble rules. Its base learners include boosted trees, ExtraTrees, random forests, Ridge regression, and Lasso regression,…
This Qlib configuration trains an ADD model using a GRU base model and Alpha360 features to rank CSI 300 stocks. Features are robustly normalized and missing values are filled; labels are cross-sectionally rank-normalized after missing labels are dropped.…
This overview describes a practical Python workflow spanning data collection, database access, and machine-learning, deep-learning, and natural-language-processing tools. Its investing example uses traditional factors measured over rolling 12-month windows…
This configuration specifies a Qlib research workflow for China A-share stocks in the CSI 300 universe. It uses Alpha360 features, robust feature normalization and missing-value filling, and cross-sectional rank normalization for labels. The prediction…
The document outlines how sequence learning can fit into a conventional quantitative investment process: feature extraction, individual asset return prediction, portfolio construction, and trade execution. It describes using price and volume measures,…
The author recounts using Kelly-style position sizing, mathematical optimization, historical backtests, and a neural network to investigate trading systems. The discussion first explains why a positive win rate alone does not prevent ruin: bet size and…
This document poses a model-design question: whether the number of input factors in a StockRank model relates to the number of leaves per tree or the total number of trees, and whether there are useful rules of thumb for choosing these quantities. It…
This article reviews Word2Vec as an efficient method for learning word vectors from context, and compares it with matrix-based and clustering approaches to distributed representations. It discusses CBOW and Skip-Gram, emphasizing their different update…
This document outlines a machine learning workflow built around feature engineering, model training, and model combination. Feature engineering includes creating, extracting, and selecting variables. It contrasts linear models, which capture linear…
The article examines a z-score indicator whose automatic lookback selection is described as machine learning. It selects among ten fixed windows using price-trend fit, then trades z-score extremes as mean-reversion signals. The audit identifies problems in…
The document explains how to choose between regression and classification when building machine-learning models for stock research. A regression task predicts a continuous quantity, illustrated by forecasting a stock’s five-day return. It introduces fitting…
The document describes an Expert Advisor that uses recent Bollinger Band Width values in a neural network method. Band Width is presented as the percentage spread between the upper and lower Bollinger Bands, normalized by the middle band in the…
The document summarizes a report that uses a random forest model to identify rising and falling market patterns from macroeconomic indicators and market data. It describes two applications: timing exposure to the CSI 300 index and rotating among industry…
The author describes a monthly stock-selection approach that buys the ten lowest price-to-book stocks in the CSI 300 universe, then combines it with a market-timing signal based on a hidden Markov model. The starting idea is a single-factor value strategy,…
The report compares ordinary K-fold cross-validation with time-series cross-validation for tuning machine-learning models on sequential data. Random folds can place later observations in training while earlier ones are used for validation, creating leakage…
This configuration describes a Qlib experiment for CSI 300 equities using a LightGBM regression model. The dataset combines daily features with intraday one-minute data resampled to a daily frequency through a custom handler. The configured label frequency…
The post describes a Chinese A-share selection rule combining membership in a metaverse-related category, high recent turnover, and a recent limit-up event. Its stated intent is to find actively traded stocks attracting attention and showing evidence of…
The tutorial outlines an end-to-end FinRL workflow for training and comparing deep reinforcement learning agents on Dow 30 equities. It describes downloading and preprocessing market data, adding technical indicators plus VIX and a turbulence measure,…
This Chinese-language strategy post describes a stock screen for the metaverse industry. It looks for shares whose opening price is near the ten-day moving average and whose recent price bars meet a proposed “arc” shape condition. The technical formula…
This advanced tutorial section is intended for readers already familiar with FinRL or its market simulation companion, or who have worked through introductory notebooks. It points to a comparison of three deep reinforcement learning libraries supported by…
This document outlines a recurring BigQuant program for people developing quantitative trading skills. It has three stages: an online selection round in which participants submit factor models built from a provided template; training for selected entrants;…