Share

D. Concepts, Systems and Data

Lectures

5

Difficulty

Easy

Homework

Yes

Small Description

Systems and data importance.

journals

•

An additional programming exercise can be found here with the recording. (code, recording). If you have any question, please email me or ask your friends on the forum. 

•

Jacques Joubert whose company works with six for the ten largest quant hedge funds gave a presentation last semester and was nice enough to share his presentation deck with us (deck, recording)

Gallery

Concepts and Systems

Reddit Sentiment

Here is the opportunity for you to create your own trading strategy, the code is built to extract some essential sentiment metrics and you have the time stamps.

4th Class Preparation Exercise

Broad Concepts

In this lecture, we will discuss some of the broad concepts common across all machine learning algorithms. These concepts provide a general framework for understanding different methods.

Specifically, we will discuss:

1. Loss functions

2. Overfitting/Underfitting

3. Feature Engineering

4. Regularization

5. The curse of dimensionality

The goal of this lecture is to explain how these concepts can be used to evaluate a given machine learning pipeline and compare different methodologies.

1. Loss Functions

In a general learning setup, we are provided with some input data, that we want to use to predict some specified output.

For any predictor that we choose to perform the above task, we can always expect the predictor to make errors in its prediction. 

This prediction error can thus be seen as a measure of performance of our machine learning algorithm. 

Furthermore, since our objective is to obtain a predictor that has the least prediction error on the set of unseen data samples, this error can be used as a signal to 'train' our machine learning algorithm. 

A. Broad Concepts

What is Alternative Data

Traditional data includes GAAP financial statements data and exchange market data that are generally mandated by law and has been used by investors for many decades. 

Alternative Data is all the data that are not mandated by law and do not fall in the above categories.

Alternative data had its start as a tool to guide investors to select the appropriate assets and stocks to buy.

It has since morphed into an industry that supplies data to companies, municipalities, policy groups, and regulators, and really anyone willing to pay the right price.

As an example, the datapoints below is not only useful for Investors, but also for:

Here you can imagine a company manager or fund manager relying on alternative data to predict revenues.

Historical Development

Data is the most important component to machine learning models, and alternative data is the most interesting form of data in finance.

Data has become especially important for finance at the turn of the century.

Renaissance Technologies’ previous CEO — and Cambridge Analytica investor — Robert Mercer was quoted saying that “There’s no data like more data”.

Their Medallion fund has averaged 72% annual return, before fees, from 1994 through mid-2014.

This applies to all financial industries, on another occasion in 2012 ZestAI’s CEO Douglas Merrill proclaimed that “all data is credit data.”

B. Alternative Data

The science of machine learning is happening inside the data and doesn’t live within an experiment, as such data collection is the lifeblood of machine learning. 

There are a few places you can go to grab some datasets: Learn how to access WRDS data, Quandl data, IEX data, Yahoo finance, Pandas data reader, Dukascopy, and other macro-economic databases.

C. Obtaining Data

Commercial Products

Marquee

Goldman Sachs’ digital marketplace for institutional investors, delivering unparalleled, cross-asset access to global markets

They have hundreds of products, the digital asset dashboard below is simply one of them. 

Marquee is essentially a platform that allows different businesses at Goldman to open themselves up to clients in a secure and consistent way. 

Before, if, say, the interest rates business wanted to share analysis with clients electronically, it would have to build something bespoke. It would be hugely time-consuming and expensive.

They seek to provide research, data, analytics, execution solutions and many more. 

Product Development

D. Developing Products

Due the 4th of November.

Web Scraping

Install required scraping libraries (can take 5 minutes). Scrape for S&P Constituents and Grab Wikipedia Page. We want to use a pre-built scraper to grab the appropriate data on the MSCI website. To learn more about the scraper you can go and visit this python repository.

Notice the scraper is broken! But I provided some code in the notebook to download the data that I scraped in the past. You don’t have to fix the scraper, that would be too hard.

Trading Strategy

After you have scraped the data, you can create a small trading strategy to see if the ESG data is helpful:

You can create a strategy that assumes that you have had access to this data since 2018.

That means that you can backtest the strategy from 2018 to 2021.

An example of the most obvious strategies is to go long and short a few companies and to measure the performance over time (long short strategy).

You can show the performance over time using a range of statistics like the sharpe ratio, but perhaps the easiest is to plot the cummulative returns.

For example the start of a very easy strategy could be:

There are many places in the internet where you can get pricing data, in fact your first homework you obtained pricing data, have a look at yfinance on github if you are still unsure.

F. Trading Strategy (HW3)

TOP