Python, Pandas and other animals to survive in the data jungle
summary
Claims an unsupervised clustering of hundreds of market features to define regime types, then trains a supervised detector to label each trading day automatically, replacing manual regime calls. Builds a 2D embedding of 37,000-dimension technical-model space, clusters model behaviors, and uses a boosting-based ranking to pick model portfolios for allocation. Implements everything in Python with Pandas, NumPy, scikit-learn, Keras/TensorFlow, LightGBM/XGBoost, Matplotlib/Seaborn and Dask; reports a recent LightGBM hackathon win. unsupervised clustering of hundreds of features automatic strategy allocation via model embedding and boosting ranking
AI-written summary · about
chapters (6)
- 0:00Introduction and background
- 8:06Python tools for data science
- 13:51Data science workflow and packages
- 17:49Machine learning tools and models
- 22:24Big data and final thoughts
- 23:48Q&A
related talks
28mUsing Python in a Data Hackathon
45mInteractive Analysis of (Large) Financial Data Sets
41mFootball's data analytics in python
33mWhy Python Has Taken Over Finance
01h 31mPandas, Data Wrangling & Data Science
51mClosing Keynote
31mBuilding Data-Driven Client Relationship Management in Banking with Python
31mPython Doesn’t Have to Be Slow: Speeding Up a Large-Scale Optimization Algorithm