Python in Finance: Actual Time Knowledge Streaming inside Jupyter Pocket book – KDnuggets #Imaginations Hub

Image source -


On this weblog, you’ll study to visualise reside information streams in actual time, all throughout the consolation of your favourite instrument, the Jupyter Pocket book. 

In most tasks, dynamic charts inside Jupyter Notebooks want guide updates; for instance, it could require you to hit reload to fetch new information to replace the charts. This doesn’t work effectively for any fast-paced trade, together with finance. Think about lacking out on essential purchase indications or fraud alerts as a result of your person didn’t hit reload at that occasion.

Right here, we’ll present you easy methods to transfer from guide updates to a streaming or real-time technique in Jupyter Pocket book, making your tasks extra environment friendly and reactive.

What’s coated:

  • Actual-Time Visualization: You will learn to convey information to life, watching it evolve second by second, proper earlier than your eyes.
  • Jupyter Pocket book Mastery: Harness the complete energy of Jupyter Pocket book, not only for static information evaluation however for dynamic, streaming information.
  • Python in Quant Finance Use Case: Dive right into a sensible monetary utility, implementing a broadly utilized in finance with real-world information.
  • Stream Knowledge Processing: Perceive the foundations and advantages of processing information in real-time, a talent changing into more and more essential in as we speak’s fast-paced information world.

By the top of this weblog, you will know easy methods to construct comparable real-time visualizations just like the one beneath inside your Jupyter Pocket book.


Python in Finance: Real Time Data Streaming within Jupyter Notebook



On the coronary heart of our mission lies the idea of stream processing. 

Merely put, stream processing is about dealing with and analyzing information in real-time because it’s generated. Consider it like Google Maps throughout a rush hour drive, the place you see visitors updates reside, enabling instant and environment friendly route adjustments.

Apparently, based on the CIO of Goldman Sachs on this Forbes podcast, shifting in direction of stream or real-time information processing is among the vital traits we’re headed towards. 



It’s about combining the facility of real-time information processing with an interactive and acquainted setting of Jupyter Notebooks. 

Moreover that, Jupyter Notebooks play effectively with containerized environments. Due to this fact, our tasks aren’t simply caught on native machines; we are able to take them anyplace – operating them easily on something from a colleague’s laptop computer to a cloud server.



In finance, each second counts, whether or not for fraud detection or buying and selling, and that’s why stream information processing has turn out to be important. The highlight right here is on Bollinger Bands, a instrument useful for monetary buying and selling. This instrument includes:

  • The Center Band: This can be a 20-period shifting common, which calculates the typical inventory worth over the previous 20 durations (akin to 20 minutes for high-frequency evaluation), giving a snapshot of current worth traits.
  • Outer Bands: Positioned 2 commonplace deviations above and beneath the center band, they point out market volatility – wider bands recommend extra volatility, and narrower bands, much less.


Python in Finance: Real Time Data Streaming within Jupyter Notebook


In Bollinger Bands, probably overbought situations are signaled when the shifting common worth touches or exceeds the higher band (a cue to promote, usually marked in pink), and oversold situations are indicated when the value dips beneath the decrease band (a cue to purchase, sometimes marked in inexperienced).

Algo merchants normally pair Bollinger Bands with different technical indicators. 

Right here, we made a necessary tweak whereas producing our Bollinger Bands by integrating buying and selling volumes. Historically, Bollinger Bands don’t take into account buying and selling quantity and are calculated solely based mostly on worth information.

Thus, we have now indicated Bollinger Bands at a distance of VWAP ± 2 × VWSTD the place: 

  • VWAP: A 1-minute volume-weighted common worth for a extra volume-sensitive perspective.
  • VWSTD: Represents a targeted, 20-minute commonplace deviation, i.e., a measure of market volatility.

Technical implementation:

  • We use temporal sliding home windows (‘pw.temporal.sliding’) to investigate information in 20-minute segments, akin to shifting a magnifying glass over the information in actual time.
  • We make use of reducers (‘pw.reducers’), which course of information inside every window to yield a selected final result for every window, i.e., the usual deviations on this case.



  • Supplier of real-time and historic market information. Whilst you can actually use its API for reside information, we have pre-saved some information right into a CSV file for this demo, making it straightforward to observe without having an API key.
  • Pathway: An open-source Pythonic framework for quick information processing. It handles each batch (static) and streaming (real-time) information. 
  • Bokeh: Preferrred for creating dynamic visualizations, Bokeh brings our streaming information to life with partaking, interactive charts.
  • Panel: Enhances our mission with real-time dashboard capabilities, working alongside Bokeh to replace our visualizations as new information streams are available.



This includes six steps:

  1. Doing pip set up for related frameworks and importing related libraries.
  2. Fetching pattern information
  3. Organising the information supply for computation
  4. Calculating the stats important for Bollinger Bands
  5. Dashboard Creation utilizing Bokeh and Panel
  6. Hitting the run command


1. Imports and Setup


First, let’s shortly set up the required parts.

%%seize --no-display
!pip set up pathway


Begin by importing the required libraries. These libraries will assist in information processing, visualization, and constructing interactive dashboards.

# Importing libraries for information processing, visualization, and dashboard creation

import datetime
import bokeh.fashions
import bokeh.plotting
import panel
import pathway as pw


2. Fetching Pattern Knowledge


Subsequent, obtain the pattern information from GitHub. This step is essential for accessing our information for visualization. Right here, we have now fetched Apple Inc (AAPL) inventory costs.

# Command to obtain the pattern APPLE INC inventory costs extracted by way of Polygon API and saved in a CSV for ease of assessment of this pocket book.

%%seize --no-display
!wget -nc


Notice: This tutorial leverages a showcase printed right here


3. Knowledge Supply Setup


Create a streaming information supply utilizing the CSV file. This simulates a reside information stream, providing a sensible strategy to work with real-time information with out necessitating an API key whereas constructing the mission for the primary time. 

# Making a streaming information supply from a CSV file

fname = "ticker.csv"
schema = pw.schema_from_csv(fname)
information = pw.demo.replay_csv(fname, schema=schema, input_rate=1000)

# Uncommenting the road beneath will override the information desk outlined above and swap the information supply to static mode, which is useful for preliminary testing
# information =, schema=schema, mode="static")

# Parsing the timestamps within the information

information = information.with_columns(t=information.t.dt.utc_from_timestamp(unit="ms"))


Notice: No information processing happens instantly, however on the finish once we hit the run command.


4. Calculating the stats important for Bollinger Bands


Right here, we are going to briefly construct the buying and selling algorithm we mentioned above. We’ve got a dummy stream of Apple Inc. inventory costs. Now, to make Bollinger Bands, 

  1. We’ll calculate the weighted 20-minute commonplace deviation (VWSTD)
  2. The 1-minute weighted operating common of costs (VWAP)
  3. Be a part of the 2 above.
# Calculating the 20-minute rolling statistics for Bollinger Bands

minute_20_stats = (
    .cut back(
        transact_total=pw.reducers.sum(pw.this.quantity * pw.this.vwap),
        transact_total2=pw.reducers.sum(pw.this.quantity * pw.this.vwap**2),
    .with_columns(vwap=pw.this.transact_total / pw.this.quantity)
        vwstd=(pw.this.transact_total2 / pw.this.quantity - pw.this.vwap**2)
        ** 0.5
        bollinger_upper=pw.this.vwap + 2 * pw.this.vwstd,
        bollinger_lower=pw.this.vwap - 2 * pw.this.vwstd,
# Computing the 1-minute rolling statistics

minute_1_stats = (
    .cut back(
        transact_total=pw.reducers.sum(pw.this.quantity * pw.this.vwap),
    .with_columns(vwap=pw.this.transact_total / pw.this.quantity)
# Becoming a member of the 1-minute and 20-minute statistics for complete evaluation

joint_stats = ( a part of(
        pw.left.t == pw.proper.t,
        pw.left.ticker == pw.proper.ticker,
        is_alert=(pw.this.quantity > 10000)
        & (
            (pw.this.vwap > pw.this.bollinger_upper)
            | (pw.this.vwap  pw.this.bollinger_upper, "promote", "purchase"
alerts = joint_stats.filter(pw.this.is_alert)


You may take a look at the pocket book right here for a deeper understanding of the computations. 


5. Dashboard Creation


It is time to convey our evaluation to life with a Bokeh plot and Panel desk visualization. 

# Operate to create the statistics plot

def stats_plotter(src):
    actions = ["buy", "sell", "hold"]
    color_map = bokeh.fashions.CategoricalColorMapper(
        elements=actions, palette=("#00ff00", "#ff0000", "#00000000")

    fig = bokeh.plotting.determine(
        title="20 minutes Bollinger bands with final 1 minute common",
        y_range=(188.5, 191),
    fig.line("t", "vwap", supply=src)
    band = bokeh.fashions.Band(
        coloration="area": "motion", "remodel": color_map,
    return fig

# Combining the plot and desk in a Panel Row

viz = panel.Row(
    joint_stats.plot(stats_plotter, sorting_col="t"),
        pw.this.ticker, pw.this.t, pw.this.vwap, pw.this.motion
    ).present(include_id=False, sorters=["field": "t", "dir": "desc"]),


Once you run this cell, placeholder containers are created in your pocket book for the plot and desk. They will be crammed with reside information as soon as the computation begins. 


6. Working the Computation


All of the preparations are full, and it is time to run the information processing engine.

# Command to begin the Pathway information processing engine
%%seize --no-display


Because the dashboard updates in real-time, you will see how the Bollinger Bands set off actions – inexperienced markers for getting and pink for promoting, usually at a barely larger worth. 


Notice: It is best to manually run after the widget is initialized and visual. You could find extra particulars on this GitHub concern right here.



On this weblog, we perceive Bollinger Bands and take you thru a journey of visualizing real-time monetary information in Jupyter Pocket book. We confirmed easy methods to remodel reside information streams into actionable insights utilizing Bollinger Bands and a mix of open-source Pythonic instruments.

The tutorial supplies a sensible instance of real-time monetary information evaluation, leveraging open supply for an end-to-end answer from information fetching to interactive dashboarding. You may create comparable tasks by:

  • Doing this for a inventory of your selection by fetching reside inventory costs from APIs like Yahoo Finance, Polygon, Kraken, and so forth.
  • Doing this for a gaggle of your favourite shares, ETFs, and so forth. 
  • Leveraging another buying and selling instrument other than Bollinger Bands.

By integrating these devices with real-time information inside a Jupyter Pocket book, you’re not simply analyzing the market however experiencing it because it unfolds. 

Pleased streaming!

Mudit Srivastava works at Pathway. Previous to this, he was a founding member of AI Planet and is an energetic neighborhood builder within the area of LLMs and Actual-time ML.

Related articles

You may also be interested in