Dr. Ernest P.Chan & Dr. Roger Hunter – Data & Feature Engineering for Trading download, Dr. Ernest P.Chan & Dr. Roger Hunter – Data & Feature Engineering for Trading review, Dr. Ernest P.Chan & Dr. Roger Hunter – Data & Feature Engineering for Trading free
Dr. Ernest P.Chan & Dr. Roger Hunter – Data & Feature Engineering for Trading
Data & Feature Engineering for Trading
AUTHORS
Dr. Ernest P. Chan
Dr. Roger Hunter
LEVEL
Intermediate
How many times have you created a strategy that performed well during backtesting, however failed to make money in the real markets? An essential course to create robust machine learning strategies which can be executed on trading platforms. This course teaches the data cleaning aspects on financial datasets and with real-world examples.
SKILLS COVERED
Data Engineering
Financial data cleaning
Exploratory data analysis
Data types nuances
Survivorship & Look ahead Bias
Feature Engineering
Triple barrier method
Dollar and volume bars
Stationarity
Fractional differentiation
Python
Itertools
Numpy
Pandas
Matplotlib
Pickle
COURSE FEATURES
Lifetime Access to the course
Downloadable codes
Video based course
Sample Strategy for Live Trading
LEARNING TRACK
Machine Learning Strategy Development and Live Trading
PREREQUISITES
You should be familiar with basic machine learning principles such as train and test datasets. There are no prerequisites as such and anyone who is familiar with financial markets data can enroll in the course.
AFTER THIS COURSE YOU’LL BE ABLE TO Preprocess price data to resolve outliers, duplicate values, multiple stock classes, survivorship bias, and look-ahead bias issues.
Work with sentiment data to identify structural break and aggregate categorical features.
Examine fundamental data and resolve multiple data merging issues.
Create features and target variables for machine learning models.
Explain various challenges associated with the financial data
SYLLABUS
Introduction to the Course
In this introductory section, you will learn the importance of data engineering and feature engineering which can be used either in your personal trading or in an institutional setting. Preprocessing of the financial dataset is essential to make it suitable for analysis. Extracting features from the datasets to feed into the machine learning algorithms, and setting the target variable for a particular ML problem increases the predictive power of your algorithm.
Introduction by Dr. Ernest Chan
4m 35s
Course Overview
4m 22s
Quantra Features and Guidance
2m 25s
Challenges in Financial Data Engineering
Most of the time, trading strategies look great while backtesting but fail to live up to the expectations during live practice. Incorrect financial data has the potential to produce inaccurate inferences. Failure in identifying the flaws in data makes it completely useless. Learn the six most common challenges in financial datasets.
Challenges in Financial Data Engineering
3m 8s
Survivorship Bias
2m
Alternative Data
2m
Exploratory Data Analysis in Finance
Exploring the data helps to build familiarity with the data. After exploring the data, you will be able to describe what’s in the data and the characteristics of the data. It also helps you to identify the irregularities and anomalies and to discover the patterns and relationships in the data.
Closer Look At the Data
2m 4s
Importance of EDA
2m
Python Pickle
2m
Adjusted Close Price
2m
How to Use Jupyter Notebook?
1m 54s
Examining the OHLCV Data
10m
Read a Pickle File
5m
Find Null Values
5m
Generate Descriptive Statistics
5m
Irregularities
2m 56s
Stock Classes
2m
Minimum Value of Adjusted Close
2m
Dataframe Profiling
2m
Survivorship Bias for Stock Data
We often backtest on the stock universe that survived until today and ignoring the stocks that no longer exists. This causes survivorship bias in the backtesting. In this section, you will learn the concept of survivorship bias, why it is important to use survivorship bias-free data in the backtesting, and how to deal with it. Also, learn to identify delisted stocks from the stock universe.
Survivorship Bias
2m 55s
Stock Disappearance
2m
Dealing With Survivorship Bias
2m
Buy-Low Price Strategy
2m
Effects of Survivorship Bias
2m
Delisted Stocks
10m
Maximum Date for Each Symbol
5m
List of Delisted Stocks
5m
Redundant Stocks Data
Learn to check for data redundancy. It is highly unlikely that two stocks or financial instruments will have the same prices across many dates. It can occur on a few dates coincidentally, but if it occurs across many numbers of dates and consecutively then something might be wrong with the data.
Dealing With Redundant Stocks
2m 43s
Effects of Redundant Data
2m
Steps to Find Redundant Data
2m
Handling Duplicate Stock Data
10m
Create Stock Pairs
5m
Compare the Stock Prices
5m
Calculate Number of Duplicates
5m
Reasons for Redundancy
2m
Multiple Stock Classes: One or All?
A listed company can issue stock with multiple classes. These stock classes have different voting rights. Learn whether you should keep the data for all the stock classes or one. If one then which stock class to keep and which to remove.
Dealing With Multiple Stock Classes
2m 1s
One Stock Class
2m
Stock Class
2m
Retain All Classes
2m
Multiple Stock Classes
10m
Identify Stock Classes
5m
Unique Symbols
5m
Outliers: How to Identify and Deal With Them?
In this section, we talk about the outliers. An outlier is a data point that is significantly different from other data points. It can be due to data quality issues or can be real. Learn how to identify and deal with outliers.
Dealing With Outliers
Outliers
2m
Dealing With Outliers
2m
Inflated Profits
2m
Dealing With Outliers
10m
Number of Trading Days With Zero Volume
5m
Sort Dataframe by Returns
5m
News Data: Numerical Features
This section covers how news data can be sourced, within the notebook, via webhose.io
Overview of the News Data
1m 20s
Numerical Features
2m 34s
Relevance
2m
Novelty
2m
Combine Numerical Features
2m 15s
Combine Numerical Features
2m
Calculate Feature Score
2m
Aggregate News Items Daily
2m
Numerical Features
10m
Calculate Feature Score
5m
Calculate Trading Date for Each Headline
5m
Calculate Daily Feature Score
5m
News Data: Categorical Features
Categorical Features
9s
Categorical Features
2m
One-Hot Encoding
2m
Aggregating Categorical Attributes
2m 16s
Aggregate Categorical Features
2m
Issues With Mean Aggregation
2m
Limitations of One-Hot Encoding
2m
Aggregating Categorical Features
10m
One-Hot Encoding
5m
Aggregate Using Mean
5m
Recap
1m 44s
Structural Breaks in Financial Data
Sometimes there is an unexpected and prolonged change in the structure of the time-series data. This leads to a structural break. Learn to identify structural breaks in the sentiment data and list the probable solutions to deal with that.
Structural Breaks
2m 48s
Structural Breaks in Time Series Data
2m
Dealing With Structural Breaks
2m
Effects of Structural Breaks
2m
Fundamental Data: Merge Them Correctly
This section covers the merging of fundamental data of two popular data sources, sharader and WSH. Although these sources are not free, the notebook also elaborates on what the data looks like and how to parse it.
Precap of Fundamental Data
58s
Sources of Fundamental Data
3m 18s
Sharadar Data
2m
Announcement and Filing Date
2m
Actual Vs Expected Earnings Date
2m
Sharadar Data
10m
Dimension Fields
2m
Why Dimension Fields?
2m
Wall Street Horizon Data
10m
Which Format?
2m
Examining the Data
2m 18s
Challenges in the Datasets
2m
Identify the Issues
2m
Multiple EPS Values
2m
Challenges in Merging Dataset
1m 5s
Common Tickers
2m
Investigate the Issues
2m
Look-ahead Bias: Deceptive Returns
Get introduced to the issues of and scenarios where data from the future is used for backtesting. This leads to deceptive returns while testing. Learn about ways to get around this ubiquitous bias or problem.
Futures Prerequisite
10m
Futures Contract
2m
Margin Requirements
2m
Settlement Price
2m
Roll Return
2m
Calculate the Roll Returns
2m
Look-ahead Bias in Futures
What is Look Ahead Bias?
2m
Good Results
2m
Futures’ Mean
2m
Remove Bias
2m
Calendar Spread Strategy Prerequisite
10m
Calendar Spread
2m
Disadvantages of CS Strategy
2m
Look-ahead Bias in CS Strategy
Problem With Two Instruments
2m
Solving the Problem
2m
Illiquid Futures
2m
Bid-Ask Time Quote
2m
Liquid Futures
2m
Types of Bars: Features Extraction
The market transaction data can be sampled in a variety of ways. For example, time, number, volume and value of transactions are different data features that can be used. But some ways might be more useful than others. Get introduced to the criteria which can be used to sample the transaction data. Also, learn about how these bars differ in their statistical properties.
Tick and Time Bars
True for a Bar
2m
Difference Between Time and Tick Bar
2m
Limitation of Time Bar
2m
Creating Time Bars
10m
Resample Price Data
5m
Calculate Open Price of Time Bar
5m
Calculate Total Volume of Bars
5m
Creating Tick Bars
10m
Aggregate Price Data
5m
Aggregate Volume Data
5m
Volume Bars
2m
Limitations of Volume Bars
2m
Volume Bar
2m
High Value of Volume Bar
2m
Limitation of Tick Bar
2m
Creating Volume Bars
10m
Create New Group ID
5m
Dollar Bars
1m 48s
What Are Dollar Bars?
2m
Identical Bars
2m
Advantages of Dollar Bars
2m
Creating Dollar Bars
10m
Information Bars: Market Order Imbalances
In this section, you will get introduced to some of the advanced ways used to sample transaction data based on market order imbalances. You will also learn market imbalances and run bars and it’s implementation.
Information Bars
2m 22s
Measure of Information
2m
Imbalance Bar
2m
Difference Between Run and Tick Bars
2m
Imbalance Bars
10m
Calculate Rolling Imbalance
5m
Additional Reading
10m
Data Labelling for Better Outcomes
Supervised machine learning algorithms need either of the two, input and a label to learn nuances of real data. In financial time series, the input is generally a window of price data. Whereas, ground truth or labels need to be explicitly generated based on the position that needs to be taken. Learn various methods like fixed time-horizon and triple barrier methods that can be used to label your data.
Fixed-Time Horizon
3m 28s
ML Paradigm Labelling
2m
Labelling Fixed Threshold
2m
The Fixed-Time Horizon Method
10m
Calculate Future Returns
5m
Labelling the Target Class
2m
Calculate Daily Returns
5m
Calculate Rolling Standard Deviation
5m
Triple Barrier Method
Fixed-Horizon V/s Triple-Barrier
2m
Calculating Horizontal Bars
2m
Horizontal Bars and Volatility
2m
Finding the Target Class
2m
Vertical Bar
2m
The Triple Barrier Method
10m
Calculate Daily Returns
5m
Call Triple Barrier Method
5m
Why Stationary Features?
The right input into a machine learning model can make all the difference in the world. Learn about the need for stationary features. Decipher the price level information vs stationarity tradeoff. Learn about fractional differentiation to create effective features.
Dealing With Features Selection
3m 10s
Price Series
2m
Series Stationarity
2m
Adjusted Close Price
2m
Fractional Differentiation
10m
Calculating Binomial Distribution Weights
2m
Calculate the ADF Statistics
5m
Python Installation
Learn to install the Python environment in your local machine.
Troubleshooting Tips For Setting Up Environment
10m
How to Run Files in Downloadable Section?
10m
Troubleshooting For Running Files in Downloadable Section
10m
Summary
This section consists of the summary of the course along with the downloadable files which include the data modules as well as the strategy notebooks.
Summary
2m 50s
Downloadable Code
2m
ABOUT AUTHOR
Dr. Ernest P.Chan
Dr. Ernest Chan is the Managing Member of QTS Capital Management, LLC., a commodity pool operator and trading advisor. QTS manages a hedge fund as well as individual accounts. He has worked in IBM human language technologies group where he developed natural language processing system which was ranked 7th globally in the defense advanced research project competition. He also worked with Morgan Stanley’s Artificial intelligence and data mining group where he developed trading strategies.
Dr. Roger Hunter
Dr. Roger Hunter is the Chief Technology Officer of QTS. He is responsible for designing high performance automated execution system that achieved negative slippage. Roger is a serial entrepreneur, having founded profitable hedge funds and software firms. Roger was formerly professor of mathematics at New Mexico State University, and he obtained his Ph.D. in Mathematics from Australian National University.
WHY QUANTRA?
Gain more in less time
Get taught by practitioners
Learn at your own pace
Get data & strategy models to practice on your own
Frequently Asked Questions:
Innovative Business Model:
Embrace the reality of a genuine business! Our approach involves forming a group buy, where we collectively share the costs among members. Using these funds, we purchase sought-after courses from sale pages and make them accessible to individuals facing financial constraints. Despite potential reservations from the authors, our customers appreciate the affordability and accessibility we provide.
The Legal Landscape: Yes and No:
The legality of our operations falls into a gray area. While we lack explicit approval from the course authors for resale, there’s a technicality at play. When procuring the course, the author didn’t specify any restrictions on resale. This legal nuance presents both an opportunity for us and a boon for those seeking budget-friendly access.
Quality Assurance: Unveiling the Real Deal:
Delving into the heart of the matter – quality. Acquiring the course directly from the sale page ensures that all documents and materials are identical to those obtained through conventional means. However, our differentiator lies in going beyond personal study; we take an extra step by reselling. It’s important to note that we are not the official course providers, meaning certain premium services aren’t included in our package:
No coaching calls or scheduled sessions with the author.
No access to the author’s private Facebook group or web portal.
No entry to the author’s exclusive membership forum.
No direct email support from the author or their team.
We operate independently, aiming to bridge the affordability gap without the additional services offered by official course channels. Your understanding of our unique approach is greatly appreciated.
Refund is acceptable:
Firstly, item is not as explained
Secondly, Item do not work the way it should.
Thirdly, and most importantly, support extension can not be used.
Thank you for choosing us! We’re so happy that you feel comfortable enough with us to forward your business here.
Reviews (0)
Reviews
There are no reviews yet.
Be the first to review “Dr. Ernest P.Chan & Dr. Roger Hunter – Data & Feature Engineering for Trading” Cancel reply
Reviews
There are no reviews yet.