Sophon Base
Transwarp Data Science Platform
Distributed Algorithms
/
Visual Modelling
/
Code Modelling
Product Introduction
Sophon Base, as an enterprise-class one-stop data science platform, supports data management, visual modeling, programmatic modeling, model management, task flow scheduling and model services, covering the full lifecycle process of machine learning models from data access, data pre-processing, to model training, model deployment, monitoring and operation and maintenance, helping enterprise customers to realize the implementation of artificial intelligence industry.
What Can Sophon Base Offer?
Cloud-native Modeling Services
It provides massive model services with the characteristics of agility, reliability, scalability, high elasticity, fault recovery, continuous updates without business interruption, etc. to ensure users' model application. At the same time, it supports fine management of model services, which is convenient for users to publish or subscribe the model services.
Distributed Algorithm Capabilities
It provides a variety of distributed machine learning operators to greatly improve the performance of the algorithms and easily handle application scenarios with high complexity, low latency, and high-volume features. At the same time, it provides diversified experimental scenario templates and one-stop interface operation, which guides users to quickly create business experiments related to business scenarios.
Various Ways to Access the Data
Relying on Transwarp's big data platform and cloud platform, it provides modules for data collection, data cleaning and data aggregation, and supports access to RDBS, HDFS, ORC, Parquet, local files and other data sources.
Visual Modelling Services
It enables users to use low code and drag-and-drop methods to quickly build machine learning models, and provides powerful modeling capabilities covering the whole process of data analytics including data access, ETL, feature engineering, model training, model application, model evaluation and model iteration, which effectively lowers the threshold of platform use.
Progressive Model Iteration
It provides cycle management of ETL processing, model training, model launch and other experiments and task flow, which can help users control the frequency of model iterations, and complete the regular iteration of the model. At the same time, it joins container model launching system to help users rolling release and horizontal capacity expansion more easily.
Why Should You Choose Sophon Base?
Low-threshold Machine Learning Modelling
Sophon Base provides drag-and-drop modeling services and recommended modeling services for the whole process of data analytics. Its excellent ETL processing ability, large number of high-performance operators and one-stop interface operation greatly lower the modeling threshold, enabling ordinary data analysts and business employees in the enterprise to quickly use Machine Learning better.
Efficient Scheduling of Modeling Tasks
Sophon Base delivers fusion computing and targeted optimization on hardware that supports various algorithms, greatly improving the computing performance and realizing flexible scheduling of enterprise modeling. At the same time, it also provides task flow scheduling service and business scenario template service, which can manage and optimize multiple experiment-triggering logic and scheduling dependencies in various business scenarios for enterprise customers.
Strong Ability of Operation, Maintenance and Management
With its enterprise level micro-service architecture, Sophon Base supports automatic service deployment and scheduling, provides powerful service scheduling and resource management and helps users with cluster operation and maintenance and security control. In this way, Sophon Base can realize the rapid deployment, rapid expansion and seamless connection of enterprise business scenario applications, and optimize the use of enterprise hardware resources.
Improving the Efficiency of User Collaboration
Sophon Base enables enterprises to share datasets, experiments, codes and models used and established in various business scenarios, and supports team collaboration, which greatly improves the efficiency of the whole team and eliminate a lot of repetitive work. At the same time, it forms a closed loop of enterprise business system docking, which can greatly promote enterprise intelligence and production applications.
Customer Cases
Customer needs
○ To achieve the goal of intelligent cabin control for 200 domestic flights, the cabin control system needs to automatically collect data from internal and external systems such as EDW, PSDP and MUFare, adopt the most reasonable cabin control strategy after demand forecasting, and finally output the seat control results to the big intelligence platform before releasing them to the subsequent business systems.
Solutions
○ Sophon Base data science platform is used to integrate various data sources such as HBase, SQL Server, Oracle, MySQL, ArgoDB and Inceptor.
○ By using the built-in high-performance distributed data mining algorithm and visual low code modeling, the historical route analysis of the operational route construction is carried out.
○ Based on Sophon Discover, the big data of intelligent cabin control is constructed by programming modeling and using a variety of machine learning frameworks.
Project results
The completed platform provides an integrated data service covering capacity, revenue and teamwork.
It provided a variety of service capabilities such as flight benefit estimation, aircraft replacement suggestion, seat control, ATPCO, dynamic group pricing and so on.
For 200 domestic flights, automated seat control is achieved over 90% of the time.
Transwarp, Shaping the Future Data World