Matt Motoki
Blog
Handling streaming data with seasonal or cyclic patterns is a common task in data science and machine learning. In this blog post, we discuss a simple implementation of a seasonal uniform average that serves as a solid baseline for forecasting...
Bayesian Target Encoding is a feature engineering method that leverages Bayesian principles to convert categorical variables into numeric representations. Beta Target Encoding is a specialized variant designed specifically for binary classification tasks. Beta Target Encoding provides a systematic way to...
Competitions
In my free time, I enjoy doing machine learning competitions. I am a competitions grandmaster on Kaggle.
- 2025 NASA Airport Throughput Prediction Challenge 1st Place $30,000
- 2025 MISO New Approaches to Baselining the Performance of DERs Top Prize $7,000
- 2024 Power Patterns Challenge Teammates: Elaine Hsu 1st Place $3,000
- 2024 Predict Energy Behavior of Prosumers Challenge 6th Place $5,000
- 2023 NeurIPS CityLearn Challenge (Forecasting Track) 1st Place $1,000
- 2023 Net Load Forecasting Prize 2nd Place $150,000
- 2023 Digitizing Utilities Prize Teammates: Sorapong Khongnawang, Daniel Tabas, Baosen Zhang, and Olin Lagon Phase 1 Winner $75,000
- 2022 NeurIPS CityLearn Challenge Teammates: Sorapong Khongnawang 2nd Place $5,000
- 2022 Solar Forecasting Prize Runner Up $25,000
- 2021 Recidivism Forecasting Challenge Teammates: Marifel Barbasa and Sorapong Khongnawang Prize Winner $48,500
- 2019 Great Energy Predictor III Teammates: Isamu Yamashita 1st Place $10,000
Projects
Cell Division is a strategy-based board game. Players take turns placing their cells. A player’s cells divide when they connect with each other. The score is equal to the total number of cells. The game ends when the board is full. The winner is the player with the most cells.
This is a collection of games inspired by Cell Division. If a player places a new cell adjacent to one or more of their existing cells, then their existing cell(s) extend to include the new cell. I’ve implemented a few different scoring mechanisms:
- Cell Multiplication: the score is the product of the sizes of the connected components
- Cell Maximization: the score is the size of the largest connected component
- Cell Connection: the score is the total number of connections of the cells
I helped to create the Hawaii Machine Learning Meetup with the goal of increasing ML adoption in Hawaii. As an organizer, I grew the community to 250+ members and helped to lead networking events and ML study groups.
