SCALAR - A platform for real-time machine learning competitions on data streams

Abstract

SCALAR is a new platform for running real-time machine learning competitions on data streams. Following the intent of Kaggle, which serves as a platform for organizing machine learning competitions adapted for batch learning, we propose SCALAR as a novel platform explicitly designed for stream learning in real-time. SCALAR supports both classification and regression problems in the data streaming setting. It has been developed in Python, using state of the art open-source solutions: Apache Kafka, Apache Spark, gRPC, Protobuf, and Docker.SCALAR is a new platform for running real-time machine learning competitions on data streams. Following the intent of Kaggle, which serves as a platform for organizing machine learning competitions adapted for batch learning, we propose SCALAR as a novel platform explicitly designed for stream learning in real-time. SCALAR supports both classification and regression problems in the data streaming setting. It has been developed in Python, using state of the art open-source solutions: Apache Kafka, Apache Spark, gRPC, Protobuf, and Docker.