{"id":6979,"date":"2026-10-25T12:37:08","date_gmt":"2026-10-25T12:37:08","guid":{"rendered":"https:\/\/eodhd.com\/financial-academy\/?p=6979"},"modified":"2026-10-25T12:37:09","modified_gmt":"2026-10-25T12:37:09","slug":"detecting-anomalies-across-correlated-assets-with-anomalybert-in-python","status":"publish","type":"post","link":"https:\/\/eodhd.com\/financial-academy\/stocks-data-analysis-examples\/detecting-anomalies-across-correlated-assets-with-anomalybert-in-python","title":{"rendered":"Detecting Anomalies Across Correlated Assets with AnomalyBERT in Python"},"content":{"rendered":"\n

A worked example using EODHD data and a self-supervised Transformer for multivariate time-series anomaly detection.<\/p>\n\n\n\n

A typical anomaly-detection example in finance looks at a single price series in isolation \u2014 a spike, a gap, a sudden jump in volatility. But some of the most useful signals only show up in the relationship<\/em> between two or more series: two normally co-moving stocks suddenly diverging, a hedge that stops hedging, a spread that breaks its historical pattern.<\/p>\n\n\n\n

This post introduces AnomalyBERT<\/strong>, the self-supervised Transformer for multivariate time-series anomaly detection proposed in Jeong et al. (2023), and walks through a small, illustrative application: computing an anomaly-score time series for the pair Coca-Cola (KO)<\/strong> and PepsiCo (PEP)<\/strong>, using daily price data pulled through the EODHD Python API.<\/p>\n\n\n\n

Register & Get Data<\/span><\/a><\/p>\n\n\n\n\n\n\n

What AnomalyBERT actually does<\/h2>\n\n\n\n

AnomalyBERT is trained without labeled anomalies<\/strong>. Instead, it learns by corrupting its own input: during training, random segments of a time-series window are replaced with one of four synthetic \u201coutlier\u201d patterns \u2014 a blended segment, a flattened segment, a lengthened or shortened segment, an injected spike \u2014 and the model is trained to recognize which timesteps were corrupted. Because the model never sees real labels, this works on any multivariate series you can supply, with no annotated anomaly dataset required.<\/p>\n\n\n\n

Input.<\/strong> A window of multivariate time series<\/p>\n\n\n\n

\"Input<\/p>\n\n

where \"N: is the window length and \"D: the number of channels. The window is split into patches of consecutive timesteps; a linear embedding layer combines the channels within each patch, and a Transformer encoder then applies self-attention across the sequence of patches (that is, across time).<\/p>\n\n

Output.<\/strong> A per-timestep anomaly score<\/p>\n\n

\"Anomaly<\/p>\n\n

produced by averaging predictions from overlapping sliding windows across the full series. A higher score means the point looked\nmore \u201ccorrupted\u201d in the sense the model learned during training \u2014 in practice, more unusual relative to the patterns\n(including the cross-series relationship) that dominate the training data.<\/p>\n\n\n\n

Note on pretrained checkpoints.<\/strong> The official implementation<\/a> ships checkpoints trained on standard industrial benchmarks (SMAP, MSL, SWaT, WADI). Those don\u2019t transfer to financial data \u2014 applying the method to a new pair of series, as below, means training your own (small) model on that series.<\/p>\n\n\n\n

1. Use the \u201cdemo\u201d API key to test our data from a limited set of tickers without registering: AAPL.US | TSLA.US | VTI.US | AMZN.US | BTC-USD | EUR-USD. Real-Time Data and all of the APIs (except Bulk) are included without API calls limitations with these tickers and the demo key.
2. Register to get your free API key (limited to 20 API calls per day) with access to: End-Of-Day Historical Data<\/a> with only the past year for any ticker, and the List of tickers per Exchange<\/a>.
3. To unlock your API key, we recommend to choose the subscription plan which covers your needs.<\/p>\n\n\n\n

Register & Get Data<\/span><\/a><\/p>\n\n\n\n

Pulling paired data via the EODHD API<\/h2>\n\n\n\n

We pull daily adjusted-close prices for KO and PEP going back to 2018, using EODHD\u2019s \/eod<\/a> <\/strong>endpoint. Note that this example needs a subscription plan with full end-of-day history: the demo key covers only a few tickers and the free key only the past year, so neither will work for KO.US and PEP.US from 2018:<\/p>\n\n\n\n

\n
\n \n
import os\nimport requests\nimport pandas as pd\n\ndef fetch_eod(ticker, start=\"2018-01-01\", end=None):\n    token = os.getenv(\"EODHD_API_TOKEN\")\n    url = f\"https:\/\/eodhd.com\/api\/eod\/{ticker}\"\n    params = {\"api_token\": token, \"period\": \"d\", \"fmt\": \"json\", \"from\": start}\n    if end:\n        params[\"to\"] = end\n    resp = requests.get(url, params=params, timeout=30)\n    resp.raise_for_status()\n    df = pd.DataFrame(resp.json())\n    df[\"date\"] = pd.to_datetime(df[\"date\"])\n    return df.set_index(\"date\")\n\nko  = fetch_eod(\"KO.US\")\npep = fetch_eod(\"PEP.US\")<\/code><\/pre>\n\n                <\/div>\n                
\n