Oskar Handmark and Michal Stypa at Backtick Technologies will share their solution to the Kaggle competition: ”Two Sigma: Using News to Predict Stock Movements”.
Kaggle competitions allow companies to explore datasets in a new way, with fresh eyes from the community. They challenge the community to find scientific solutions for a specific task. The community participates from sheer interest and a chance to win prize money.
We will go through the competition format of Kaggle, some things to think about and watch out for, using the featured competition: ”Two Sigma: Using News to Predict Stock Movements” as an example. We will cover kernels, commits, the Kaggle docker container, submissions, strategies and general tips.
We will do an Exploratory Data Analysis of the competition and share our entire solution with you!
This is a good opportunity to get into statistical models for stock market predictions. Practical application of the following subjects will be covered: Time series analysis, validating time series, working around memory limitations, K-Means clustering, Lightgbm, voting ensembles, financial indicators, lag features, binary and multiclass classification, custom eval metrics, sharpe ratio and more.
After the presentation, there will be a hands on session where you get the chance to get started on your own solution for the competition problem. We will share a few different kernels of varying difficulties that you can base your work on. This part requires that you bring your own computer, a kaggle account, intermediate knowledge of python, pandas, numpy and general ML techniques.
17.30 - 17.45 Meet & Greet
17.45 - 18.20 Presentation
18.20 - 18.30 Get into HandsOn groups
18.30 - 19.00 Meet & Eat
19.00 - 20.00 Keep working on data science exercises
20.00 - Meet & Geek
Oskar Handmark & Michal Stypa both hold MSc in Computer Science and Engineering from LTH, and has been working with big data since 2016. They started Backtick Technologies in 2018, a consultancy company specializing in data science and machine learning.