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 configuration defines a Qlib equity forecasting workflow using the CSI 300 universe and the Shanghai CSI 300 index as benchmark. It sets a data window from 2008 through 2020, with training through 2014, validation over 2015–2016, and testing over…
This discussion addresses whether a trader should use hyperparameter search to find an optimal combination of existing factors. The response says that “optimal” needs a clear definition: it could mean the best backtest result or the best factor-analysis…
This overview traces major developments in deep learning from the perceptron and backpropagation through convolutional and recurrent networks, generative models, residual networks, Transformers, and BERT. It describes the basic contribution associated with…
This research applies feature selection to multi-factor stock selection in the full Chinese A-share universe, with the CSI 300 and CSI 500 as benchmarks. It constructs industry- and market-cap-neutral strategies and tests feature ranking methods with…
The document describes a stock-selection screen that combines three filters: ranking shares by capital inflow, requiring specified price-to-earnings and price-to-book ranges, and limiting the gain at 9:25 to below a stated threshold. It frames capital flows…
This overview introduces ten machine learning methods and sketches how they work, with examples of possible financial uses. It covers linear and logistic regression, K-nearest neighbors, support vector machines, decision trees, random forests, neural…
This document describes a software class for a probabilistic neural network used for classification. A constructor takes the input-vector size and number of target classes; class labels are zero-based and must be consecutive. Training data are supplied as…
This research summary describes ranking international liquid stocks with a random forest trained on technical features. Rather than forecasting exact returns, the model estimates the likelihood that a stock will outperform an average-return threshold. The…
The document describes a MetaTrader script that collects broker-specific trading costs for one symbol and exports them for a companion model used in triple-barrier labeling. It samples spreads from ticks and reports tick-weighted, dwell-time-weighted, and…
FinRL-Meta addresses a research infrastructure problem: deep reinforcement learning has potential in finance, but researchers need realistic market environments and shared benchmarks to develop and compare methods. The document contrasts this need with…
This Chinese-language discussion describes problems encountered when reusing a saved XGBoost model in a separate strategy and moving a feature-based strategy from research into simulated trading. Keeping visible strategy settings unchanged did not reproduce…
This technical report describes a dtype-sensitive error in VeighNa’s linear-decay time-series calculation using Polars rolling_map. In the reported example, a sparse binary series with one nonzero value produces zeros when stored as Int64, while the Float64…
This meetup compilation surveys practical topics in quantitative investing, including factor design, machine learning, strategy validation, portfolio sizing, live execution, and sector rotation. It recommends building a factor library by collecting and…
The document describes a library that packages deep learning models for quantitative research, including feedforward networks, LSTMs, Transformers, TabNet, and TRA. Its quick-start workflow covers selecting a model, configuring an optimizer and loss…
This overview explains machine learning as a way to model patterns in data and describes how that capability may be applied to investment decisions. It outlines a workflow spanning data collection, feature creation, preprocessing, model training, model…
The document explains how FreqAI trains trading agents through reinforcement learning. An agent processes historical candles and chooses among actions such as entering or exiting long and short positions. A custom reward function scores its decisions, while…
The article explains why standard stochastic gradient descent can remain inefficient when training Word2Vec on a large vocabulary. Even when an update concerns only a few words, calculating gradients across the full parameter matrices wastes work, and the…
This document explains an indicator that converts each bar’s timestamp into hour-of-day features for data collection and modeling. It provides 24 binary buffers, one for each hour, and an additional buffer containing the numeric hour. For each bar, the…
This overview explains XGBoost regression as an ensemble of decision trees built sequentially to reduce a loss. Each new tree uses gradient information from prior predictions, while complexity controls and regularization are intended to limit overfitting.…
A user asks why a BigQuant query for Ping An Bank returns values in the thousands when the stock trades around the teens. The supplied workflow selects the A-share instrument, then builds a forward label from the close five sessions ahead divided by the next…
The document outlines a chart indicator that trains a support vector machine (SVM) to predict buy and sell signals. It selects technical indicators as model inputs, gathers historical indicator and price data, labels past opportunities according to whether…
This article argues that training one stock-selection model on all market data may struggle with noisy observations and changing factor behavior. Its proposed approach is to classify market conditions and maintain strategies suited to different regimes:…
This post outlines an A-share stock-selection framework that uses eight cross-sectionally ranked features covering valuation, momentum and reversal, and liquidity. A LightGBM regression model predicts relative returns over a five-trading-day horizon; the…
This example outlines a long-term equity selection process inspired by fundamental value investing. A research agent reviews the latest company reports and current share prices, then selects businesses based on profitability, competitive durability, and…