Machine learning models learn from the observations they are given. OHLCV provides a useful foundation, but many financial ML workflows require additional information about trades, liquidity, order flow, timing, and market state.
FinFeedAPI provides T+1 historical OHLCV and market metadata across 40+ exchanges, with detailed Native IEX historical microstructure data for supported symbols and dates. ML teams can use these datasets for feature engineering, model training, historical evaluation, validation, and market research.
Historical price data makes it easy to calculate returns, moving averages, volatility, and other common features.
But models trained only on aggregated price series may miss information about spreads, displayed liquidity, individual trades, order book activity, trading sessions, and unusual market states.
OHLCV captures aggregated price and volume information but removes much of the activity within each period.
Without deeper datasets, feature engineering often concentrates on transformations of the same underlying price series.
Spread, depth, and order book activity require data beyond standard candles.
Training observations from different sessions or unusual trading periods may be treated as equivalent when they occurred under different conditions.
Combining datasets from different providers, schemas, and identifier systems creates additional preprocessing and data-management work.
Suppose an ML model is trained to analyze short-term stock price behavior.
OHLCV can provide historical returns, ranges, volume, and volatility. But a richer feature set might also ask:
These observations can become additional features, filters, labels, or contextual variables depending on the model and research methodology.
FinFeedAPI provides T+1 historical OHLCV across 40+ exchanges through a standardized interface. ML teams can use this data for traditional price and volume features, while supported Native IEX datasets provide additional inputs for more detailed microstructure research.
Native IEX Level 1 provides historical best bid, best ask, and associated sizes. Native IEX Level 2 provides displayed liquidity across multiple price levels. These observations can support features related to spreads, top-of-book conditions, depth, liquidity distribution, and changes in displayed liquidity.
Native IEX Level 3 provides individual historical order book lifecycle events. ML researchers can derive features from supported order additions, modifications, deletions, executions, and book-clearing activity rather than relying exclusively on aggregated prices.
Native IEX system events provide information including system hours, regular market hours, message boundaries, and session-state transitions. Administrative messages add symbol-level context including trading status, auctions, operational halts, official open/close prices, and other relevant events. These fields can help researchers segment, filter, or contextualize training observations.
Native IEX historical trades include supported condition flags such as odd lots, extended hours, intermarket sweeps, trade breaks, trade-through exemptions, and single-price crosses. ML pipelines can use these fields when defining features, filtering observations, or constructing labels.
REST and JSON-RPC provide structured access to historical Stock API resources. Teams can automate data retrieval and preprocessing across supported symbols and historical periods, making it easier to reproduce datasets across experiments.
FinFeedAPI provides the historical market data; ML teams remain responsible for feature methodology, label construction, leakage controls, model validation, data retention, and applicable usage or redistribution requirements.
Historical OHLCV can be used to calculate returns, volatility, momentum, ranges, volume changes, rolling statistics, and other common financial ML features. These datasets can provide the baseline for models across supported instruments and exchanges.
Native IEX Level 1 data can support features based on bid-ask spreads, quoted sizes, and changes in top-of-book conditions. Researchers can compare these observations with subsequent historical market activity.
Native IEX Level 2 provides displayed liquidity across price levels. ML teams can derive features describing depth, liquidity concentration, imbalance, or changes in displayed liquidity according to their own methodology.
Native IEX Level 3 events allow researchers to study individual historical order book activity. This can support experimental features based on order additions, modifications, deletions, executions, book-clearing activity, and changes in the displayed book.
System events and administrative messages can help separate observations according to session state or supported symbol-level events. Instead of automatically treating every timestamp as equivalent, researchers can explicitly define which periods belong in a training or evaluation dataset.
Standardized historical access makes it easier to define datasets using explicit symbols, dates, periods, and data types. Teams can reuse those definitions across feature experiments, model versions, and evaluation workflows.
ML teams can begin with historical OHLCV and add more detailed Native IEX datasets when the research question requires market microstructure.
| Data | What It Provides | How It Can Support ML |
|---|---|---|
| OHLCV | T+1 historical open, high, low, close, and volume across 40+ exchanges | Returns, momentum, volatility, volume, rolling statistics, and other price-based features |
| Market Metadata | Exchange and symbol information | Dataset definition, instrument discovery, and consistent data organization |
| Trades | Individual historical trades and supported condition flags | Trade-level features, filtering, and label construction |
| Level 1 | Historical best bid, best ask, and associated sizes | Spread, quoted-size, and top-of-book features |
| Level 2 | Displayed liquidity across price levels | Depth, liquidity-distribution, and imbalance features |
| Level 3 | Individual order book lifecycle events | Detailed order-flow and book-activity features |
| System Events | System hours, regular market hours, message boundaries, and session-state transitions | Session segmentation and market-state context |
| Administrative Messages | Trading status, auctions, operational halts, official open/close prices, and other symbol-level information | Filtering and contextualizing unusual historical periods |
OHLCV and market metadata provide the broadest foundation for ML workflows. Trades, Level 1, Level 2, Level 3, system events, and administrative messages described here are Native IEX historical datasets for supported symbols and dates.
| What the ML Workflow Needs | Before | With FinFeedAPI |
|---|---|---|
| Historical training data | Models may rely primarily on aggregated price series | T+1 historical OHLCV provides standardized price and volume observations across 40+ exchanges |
| Feature engineering | Features are concentrated around transformations of price and volume | Native IEX datasets can add spread, liquidity, trade, and order book variables |
| Spread information | Spread may be unavailable or approximated | Native IEX Level 1 provides historical bid/ask observations |
| Market depth | Liquidity structure is absent from OHLCV | Native IEX Level 2 provides displayed liquidity across price levels |
| Order flow | Individual order book activity is unavailable | Native IEX Level 3 provides historical order book lifecycle events |
| Trade filtering | Individual trade conditions may not be available | Supported trade flags can be used for filtering, feature engineering, or labeling |
| Market-state context | Different historical periods may be treated as equivalent | System events and administrative messages provide additional session and symbol-level context |
| Experiment reproducibility | Datasets may depend on custom scripts and multiple sources | Structured API access supports explicitly defined and repeatable historical data retrieval |
FinFeedAPI provides structured interfaces for retrieving historical data used in feature engineering, model development, historical evaluation, and validation workflows.
REST works well for research scripts, notebooks, scheduled data collection, feature-generation pipelines, and model-development environments.
Teams can request supported historical data for defined symbols and periods and process it within their own ML infrastructure.
JSON-RPC provides structured RPC-based access for supported historical stock workflows.
It can be integrated into automated data preparation, experimentation, and research pipelines.
FinFeedAPI also provides a hosted Stock Historical MCP server.
Supported tools for exchange and symbol discovery, OHLCV, and Native IEX historical resources can be used in compatible AI-assisted research and data-analysis workflows.
A basic ML workflow might begin with:
The next step is defining what information the model should actually receive.
The appropriate dataset depends on the hypothesis being tested. More data does not automatically create a better model, but it gives researchers additional market information to evaluate.
Start with T+1 historical OHLCV and add trade, spread, liquidity, order book, and market-state information when your research requires it.
FinFeedAPI provides historical OHLCV across 40+ exchanges, market metadata, and detailed Native IEX historical datasets where supported, giving ML teams a structured data foundation for feature engineering, model training, historical evaluation, validation, and experimentation.
