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
This BigQuant support exchange explains a prediction error associated with missing or infinite values while constructing label bins. It suggests checking whether the target label is absent or contains many missing values. The further diagnosis is that the…
A BigQuant forum answer addresses why a stock selection strategy’s returns changed from positive to negative after neutralizing and standardizing its inputs. The reply explains that StockRanker uses a tree-based structure and, according to the respondent,…
This study examines whether generative adversarial networks can produce realistic multi-asset financial time series. It combines the convolutional architecture of DCGAN with the Wasserstein-distance loss used by WGAN: convolutional layers replace some…
The document outlines a training program on cross-sectional multi-factor strategies, also described as alpha strategies. Its curriculum spans factor data preparation, supervised learning, model evaluation and interpretation, portfolio construction, and…
The tutorial presents a deep reinforcement learning workflow for trading a portfolio of Dow 30 stocks. It frames trading as a Markov decision process: the agent observes prices and engineered features, including MACD and RSI, then outputs per-stock actions…
This report summarizes stock-selection results for several machine-learning models across broad Chinese equity universes and constituent groups tied to the CSI 300, CSI 500, and CSI 800. It compares weekly, recent-month, and 2018-to-date excess returns, and…
This brief Q&A explains why neutralizing individual factors for market capitalization does not guarantee that a trained stock-selection model will have no size exposure. After neutralization, linear groupings of the factor may contain both large- and…
The strategy combines model-based stock ranking with trend and timing filters for a selected universe of China A-shares. The author describes ranking stocks by predicted prospects, then screening for recent relative strength, liquidity, large-order money…
This Qlib workflow configuration specifies a linear ordinary least squares model using the Alpha158 feature handler for the CSI 300 universe. The data span begins in 2008 and ends in 2020; the training segment runs through 2014, validation covers 2015–2016,…
A forum exchange asks how research conducted with vn.py’s alpha module can be connected to live trading and whether a matching live example is available. The reply says there is no open-source template in the discussion and outlines a minimal workflow: load…
This tutorial walks through a basic Skip-gram word2vec training workflow. A center word is represented by an input vector, and the model predicts each word in its context window using a softmax output layer. The article describes calculating the prediction…
This study compares six cross-validation approaches for tuning machine-learning models on financial time series. It argues that ordinary shuffled K-fold validation can leak temporal structure and select overly complex models. Expanding-window time-series…
This document is a broad glossary of machine learning and data mining terms, organized by task. It surveys sampling methods, clustering, classification and regression, model evaluation, probabilistic graphical models, neural networks, deep learning,…
This overview introduces ten supervised learning approaches: linear and logistic regression, linear discriminant analysis, decision trees, naive Bayes, k-nearest neighbors, learning vector quantization, support vector machines, bagging and random forests,…
The document summarizes research on a listwise learning-to-rank loss designed for constructing long-short stock portfolios. The proposed loss builds on ListMLE but gives extra emphasis to the top and bottom of the ranking, where long and short selections are…
The document surveys the deep reinforcement learning agents available through FinRL, which integrates implementations from ElegantRL, Stable Baselines 3, and RLlib. The listed algorithms include value-based, policy-gradient, actor-critic, and multi-agent…
This short strategy summary describes a quantitative stock-selection approach built with TabNet, a machine-learning model. It says the approach extracts 98 price and volume factors, trains the model on data from 2010 through 2018, and applies predictions to…
The document summarizes research on using corporate earnings-call transcripts to predict subsequent stock-price movements. Earnings calls contain management discussion of financial performance and information that investors and analysts may use when making…
The document outlines BBMA, a trading approach combining Bollinger Bands and moving averages, and names several signal groups: candlestick direction patterns, extreme conditions and low-volume observations, momentum and re-entry, and a zero-loss zone. Its…
The document proposes adding a simple API to retrieve selected strategy performance metrics over a specified date range. Examples include the Sharpe ratio, annualized return, benchmark return, and alpha. The request responds to a gap between an existing…
The discussion explains that a supervised model learns from factors paired with labels, where a label defines the target. For stock selection, the target might be future returns over a chosen horizon. The model and label together determine what predictions…
This brief introduction contrasts linear regression on factor data with tree-based machine learning. It argues that factor returns may relate to outcomes in nonlinear ways, making a simple linear model potentially inadequate. Decision trees are presented as…
This brief event description presents a practical application of large language models to automated trading. Its central idea is a digital trader that interprets natural-language instructions and connects them to trade execution, with examples intended to…
This event outline introduces advanced FinRL development topics, including the FinRL-Tutorials project, reinforcement-learning portfolio allocation, stock trading in China’s A-share market, and ensemble strategies. It distinguishes portfolio allocation from…