Knowledge library

This brief introduction contrasts linear regression of factor data with decision trees. Its central point is that market returns may not relate to factors in a simple linear way, so a tree-based method is worth considering when studying how factors…

This report presents an A-share stock-selection framework that applies StockRanker, a gradient-boosted learning-to-rank model, to fundamental, financial, and trading factors. It tests 282 individual factors across short, medium, and long holding horizons,…

The document explains how to adapt a StockRanker multi-factor stock-selection workflow to Hong Kong and US equities on BigQuant. The backtest uses BigTrader’s custom historical data input, which must include date, instrument, and daily open, high, low,…

This document outlines an attempt to build a trend indicator using a neural network and Kaufman’s adaptive moving average as a reference. The described initialization phase trains the network on bars numbered 200 through 300 by selecting weight coefficients.…

This overview surveys six ICLR 2022 studies: periodic time-series forecasting, model-based policy optimization, disentangled representations, Stein-based sampling, deployment-efficient reinforcement learning, and training with privileged observations. The…

This article explains why selecting useful input features matters in predictive modeling. Irrelevant variables can add training time, contribute to overfitting, and make it harder for a model to generalize. Feature selection is presented as a way to reduce…

This Chinese-language article introduces a workflow for building custom stock-selection strategies on the older BigQuant platform. It contrasts platform-generated AI strategies, which learn from selected factors but may be difficult to interpret or tune,…

This code excerpt implements neural-network components for a momentum forecasting model based on a temporal fusion transformer design. It includes feed-forward layers, gated linear units, gated residual networks with skip connections and normalization, and…

The article reviews the progression from earlier GPT language models to GPT-3, emphasizing scale and the ability to perform varied tasks from prompts or examples without task-specific fine-tuning. It describes code generation alongside broader language…

This article explains random forests and decision trees, then applies a random forest regressor to short-horizon Chinese stock selection. It describes bootstrap sampling, random feature selection, averaging tree predictions, tree pruning, regression split…

The document describes why quantitative investors use machine learning to analyze large datasets, forecast market direction, and select portfolios. It highlights the ability of these methods to represent nonlinear relationships and process data efficiently,…

This review surveys three directions in factor investing: refining established factor methods, incorporating new data and models, and using factors as a basis for asset allocation. It describes improving value signals with multiple valuation measures,…

This review surveys three directions in factor investing: refining established factor methods, incorporating new data and models, and using factors as a basis for asset allocation. It describes improving value signals with multiple valuation measures,…

This research summary describes a machine-learning approach to predicting very short-term price changes from futures order-book data. It proposes using support vector machines with features such as best bid and ask prices and sizes, order-book depth, slope…

The document outlines a data mining workflow for investment research, moving through data collection, cleaning, feature extraction, structuring, storage, analysis, and evaluation of results. It frames this workflow as a way to turn growing volumes of market…

This Qlib documentation describes a workflow for generating, storing, training, and collecting multiple research tasks. A task can include a model, dataset, and recorded outputs. Task generators such as RollingGen can create tasks for different date…

This overview discusses two research approaches to using deep learning for equity factor discovery. The first builds stock embeddings from co-investment patterns in fund portfolios: random walks on a stock-fund graph produce sequences for skip-gram training.…

This module outlines an out-of-sample forecasting workflow built around Auto-ARIMA. It first applies an Augmented Dickey-Fuller test at a five percent significance level, repeatedly differencing the training series until the test indicates stationarity or a…

This short introductory note describes quantitative investing as the use of mathematical models to turn investment ideas into measurable criteria. Those criteria can be used to identify opportunities, place trades, and manage risk. It highlights potential…

The document explains Bayesian optimization as a way to search expensive black-box objectives, such as portfolio return or risk-adjusted performance. It describes using a Gaussian process as a probabilistic model: observations update the model’s posterior,…

This article collects practical advice for building and debugging deep learning models, using a generative adversarial network for manga coloring as its motivating example. It recommends beginning with a small working baseline, visualizing predictions and…

This Freqtrade strategy framework uses sets of generated buy and sell rules, called spells, assembled from technical indicators, operators, and numeric thresholds. Its documentation describes deriving the rules through hyperoptimization, then assigning…

This meetup Q&A surveys quantitative research topics, including collecting and cleaning market data, creating features, training models, and evaluating strategies through backtests. It lists possible information sources beyond price and volume, such as…

This overview presents BERT, a language representation model pretrained with context from both the left and right sides of text. It contrasts that design with earlier unidirectional language models, which restrict each token to information from preceding…