OHLCV Anomaly Detection Tutorial¶

This notebook demonstrates how to detect anomalies in stock price data using chronos-lab.

Setup¶

Install and import required libraries and define the list of tech stock symbols to analyze.

In [ ]:

!pip install chronos-lab[yfinance,analysis,visualization]

!pip install chronos-lab[yfinance,analysis,visualization]

In [1]:

from chronos_lab.sources import ohlcv_from_yfinance
from chronos_lab.analysis.driver import AnalysisDriver
from chronos_lab.plot import plot_ohlcv_anomalies

symbols = [
    "AAPL", "MSFT", "GOOGL", "AMZN", "NVDA", "META", "TSLA", "ORCL", "AVGO", "INTC"
]

from chronos_lab.sources import ohlcv_from_yfinance from chronos_lab.analysis.driver import AnalysisDriver from chronos_lab.plot import plot_ohlcv_anomalies symbols = [ "AAPL", "MSFT", "GOOGL", "AMZN", "NVDA", "META", "TSLA", "ORCL", "AVGO", "INTC" ]

Fetch Historical Data¶

Download one year of daily OHLCV data from Yahoo Finance for the selected symbols.

In [2]:

ohlcv = ohlcv_from_yfinance(symbols=symbols, period='1y')

ohlcv = ohlcv_from_yfinance(symbols=symbols, period='1y')

[*********************100%***********************]  10 of 10 completed

Detect Anomalies¶

Run anomaly detection using Isolation Forest algorithm. This filters symbols with anomalies present within last 3 months and saves results to a dataset.

In [3]:

driver = AnalysisDriver()

driver = AnalysisDriver()

In [6]:

anomalies = driver.detect_anomalies(ohlcv, to_dataset="enabled")

anomalies = driver.detect_anomalies(ohlcv, to_dataset="enabled")

Generate Plots¶

Create Bloomberg-style candlestick charts with highlighted anomalies for each symbol, where anomalies are detected within last 3 months.

In [14]:

if 'analysis_result' in anomalies:
    symbols_to_plot = anomalies['analysis_result'].index.get_level_values('symbol').unique().sort_values().tolist()

    for symbol in symbols_to_plot:
        ohlcv_anomalies_df = anomalies['analysis_result'].xs(symbol, level='symbol', drop_level=False)
        plot_data = plot_ohlcv_anomalies(ohlcv_anomalies_df, anomaly_period_filter='3m')
else:
    print("No anomalies found.")

if 'analysis_result' in anomalies: symbols_to_plot = anomalies['analysis_result'].index.get_level_values('symbol').unique().sort_values().tolist() for symbol in symbols_to_plot: ohlcv_anomalies_df = anomalies['analysis_result'].xs(symbol, level='symbol', drop_level=False) plot_data = plot_ohlcv_anomalies(ohlcv_anomalies_df, anomaly_period_filter='3m') else: print("No anomalies found.")

2026-02-02 10:28:46,185 | root | plot_ohlcv_anomalies | WARNING:No anomalies found for symbol AAPL. Skipping plot generation.
2026-02-02 10:28:46,186 | root | plot_ohlcv_anomalies | WARNING:No anomalies found for symbol AMZN. Skipping plot generation.

2026-02-02 10:28:46,608 | root | plot_ohlcv_anomalies | WARNING:No anomalies found for symbol GOOGL. Skipping plot generation.
2026-02-02 10:28:46,609 | root | plot_ohlcv_anomalies | WARNING:No anomalies found for symbol INTC. Skipping plot generation.

2026-02-02 10:28:47,556 | root | plot_ohlcv_anomalies | WARNING:No anomalies found for symbol NVDA. Skipping plot generation.

2026-02-02 10:28:47,943 | root | plot_ohlcv_anomalies | WARNING:No anomalies found for symbol TSLA. Skipping plot generation.

Inspect Calculation Pipeline¶

Dive into the complete execution graph powering the analysis. This visualization reveals every runtime input, transformation, and dependency that drives the anomaly detection pipeline from raw data to final results.

In [9]:

from IPython.display import HTML

g = driver.drivers['detect_anomalies'].display_all_functions()

HTML(f"""
<div style="width:100%; overflow-x:auto; border:1px solid #ccc;">
    {g._repr_image_svg_xml()}
</div>
""")

from IPython.display import HTML g = driver.drivers['detect_anomalies'].display_all_functions() HTML(f"""

{g._repr_image_svg_xml()}

""")

Out[9]:

In [ ]: