One-stop Machine Learning Modelling Platform
Product Brief Introduction
Sophon MLDev, as an advanced version of Sophon Base, on the one hand, fully takes over the data management and machine learning modelling capabilities of Sophon Base; on the other hand, Sophon MLDev has also deeply transformed the underlying framework of the system to enable lightweight deployment of the product.
Sophon MLDev, as an enterprise-class one-stop machine learning modeling platform, integrates three functional modules: Sophon Data for data management, Sophon VLab for visual modeling and Sophon Discover for programming modeling, covering the entire machine learning modeling lifecycle process from data access, data pre-processing, to model training, model evaluation and model iteration. It helps enterprise customers to realize the implementation of artificial intelligence industry.
What can Sophon MLDev offer?
Interactive Data Pre-processing
Sophon VLab has a large number of general feature engineering operators built in, enabling general data analysts and business people to quickly get started processing data and building features through a visual interface. Automatic feature engineering operators are also available to automatically build higher-order combinations of features based on original features and targets.
Distributed Machine Learning
Sophon VLab includes a wide range of distributed machine learning operators, as well as support for configuring advanced parameters related to distributed computing for specific high performance self-developed algorithm operators, providing a convenient and easy-to-use model training experience while significantly reducing model validation and iteration cycles.
One-stop Model Deployment and Maintenance
Sophon VLab supports seamlessly interfacing with Sophon MLOps, Transwarp self-developed AI capability operation platform, and flexibly scale model prediction services through a containerized architecture, helping users to quickly deploy, monitor and iterate their models.
Why Should You Choose Sophon MLDev?
Multi-source Data Access
Sophon MLDev supports access to multiple data sources such as JDBC, Hive, HBase, HDFS and StarHub's own data store; it also supports data exploration and data cleansing according to a unified view and specification for the data accessed by the system, and provides various functions such as a file management system.
Support for Flexible Deployment
and Elastic Scaling
Sophon MLDev is built using a modular microservice approach, which enhances the flexibility of installation and deployment, and allows users to install VLab or Discover modules in combination or individually as needed, supporting rapid expansion and seamless interfacing. At the same time, different services can be scaled up as needed, easily coping with the differentiated system load requirements on the user side and significantly improving the efficiency of the use of hardware resources in the user cluster.
It provides visual modelling and recommended modelling services covering the whole process of data analysis, including data access, ETL, feature engineering, model training, model evaluation and model iteration, allowing users to complete machine learning modelling without writing codeand effectively reducing the threshold of use.
Simple and Efficient
With support for multiple programming languages and interactive code output, the minimalist interface and visual Dashboard overview help users get started with managing their Sophon programming modelling projectsAnd the project resource allocation service and pipeline training task management service help users achieve efficient modelling.
○ The rule experience of traditional abnormal transaction identification is not enough to support the current high-frequency transaction scenario
○ The current low intelligence of transaction behavior monitoring affects business response efficiency
○ Risk Contorl Modeling Based on Machine Learning: based on collaborative development of Sophon Vlab, conducting the rule discovery with a combination of expert rules and machine learning to build a decision making system with risk control measures
○ Abnormal Transaction Identification Assisted with Reinforcement Learning: using Sophon Discover to build an inverse reinforcement learning model, and by means of identification analysis of counterstrike transaction, identification analysis of over-the-counter capital allocation account and abnormal trading index calculation, analyzing the characteristics of the market and identifing abnormal trading modes.
Implemented a dual-track decision-making model including expert rules and AI models, enhancing the interpretability of AI models in business scenarios
After the abnormal transaction identification system built with assistance is launched, it can identify the risks of different customers more accurately, detect abnormal transactions in a more timely manner and predict market risks in advance
Transwarp, Shaping the Future Data World