Data Intelligence, at your fingertips! Kangaroo cloud of stack introduction of scientific data platform DTinsight.Science

Understand a word DTinsight.Science

DTinsight.Science is "interactive visual modeling and code written in one of the machine learning platform exploration work", data access, visualization, modeling experiment, Notebook programming, model training, model deployment, operation and maintenance tasks such as development scenarios to help enterprise build algorithms service capabilities, to provide efficient, safe and stable algorithm operating environment.

Use of stack - data science platform DTinsight.Science one-stop data exploration

Draw focus!

The number of stack - What is data science platform DTinsight.Science:
machine learning, exploration work platform

The number of stack - data science platform DTinsight.Science product objectives are:
to provide professional, reliable and efficient algorithm modeling platform, one-stop complete algorithm modeling, model training, model deployment job machine learning, intelligent data so that your fingertips .

The number of stack - data science platform DTinsight.Science customer value are:
to help companies build algorithms service capabilities, allows intelligent data quickly fall to the ground to build business intelligence and intelligent two-way data-driven engine.

The number of stack - data science platform DTinsight.Science who used to:
enterprise data scientists and data analysts

Further in-depth understanding of DTinsight.Science

2.1 R & D mind

With big data in the enterprise application of horizontal and vertical depth, has been off-line computing, real-time computing, big data products to solve customers used to calculate the operation and maintenance scenarios. At present, the intelligent application data is also more enterprise applications, explore the direction, so that the intelligent data-driven business intelligence. In this process, how to make more intelligent business data quickly have the ability to think is the direction of the data science platform.

So today, the number of stack introduction of scientific data platform to build algorithms modeled after the data processing capabilities, to provide comprehensive and easy to use development platform for data scientists Data Analyst, hoping to provide enterprises with a more scientific and accurate judgments, to achieve high-value data application.

2.2 Advantages

# Compatible with popular machine learning and deep learning framework #
TensorFlow MXNet the Spark Python

The number of stack - data science platform DTinsight.Science currently supports TensorFlow, MXNet, Python, Spark computing framework, different algorithms select the appropriate task computing framework, support multidimensional development scenarios. Follow-up will also support Pytorch, XGboost other computing framework.

image

# # Rich component library of algorithms
your component needs, I will have

The number of stack - data science platform DTinsight.Science algorithm encapsulates many types of components, including data source / destination, SQL scripts, tools, data preprocessing, feature engineering, statistical analysis, machine learning, deep learning, prediction, assessment, text analysis, network component analysis algorithm, fully covering algorithm modeling scenarios.

image
DTinsight.Science has a rich algorithmic components

# # Visual modeling experiments
visual, drag pulled way algorithm modeling

Big Data applications in depth now, more and more people to intelligent data to the business combination, the algorithm model building full of interest. The number of stack - data science platform DTinsight.Science supports visual experiments modeling can be algorithmic components drag drag to the canvas area, consisting of visual experiments stream, can interface of the configured component parameters, and visualization view data operating results and the model results, simple and easy to get started.

image
DTinsight.Science supports visual modeling experiments

# # Interactive data exploration Notebook
using Python freely explore the data

In addition to the visualization of experimental structures, scientific data along with an interactive Notebook platform development environment, code can be written in python, execute certain statements to view the results and run logs, providing data free to explore the environment for the data analyst.

image
Notebook free interactive data exploration

# # Model offline training
cycle scheduling, training model parameters

After the algorithm model development is completed, the task may be submitted offline scheduling, training scheduling engine model according to the scheduling cycle, generate new model parameters and data. After completion of the training model, you can view assessment results, select the appropriate model parameters stored application, or online deployment.

# # Model Online deployment
models and the deployment of online calls, real-time data mining value

当模型的预测及评估结果达到目标后,可进行模型在线部署,将模型部署至模型服务器上,并生成模型的调用API。部署后,可进行模型的在线调用。

除以上核心功能外,基础的租户隔离、项目创建管理、用户管理、角色管理、数据接入等也是具备的,可一站式完成机器学习作业,欢迎大家体验。

使用DTinsight.Science可以做哪些有意思和有价值的事情?

数栈-数据科学平台DTinsight.Science可帮助企业构建算法服务能力,服务于企业数据的多样化应用场景,驱动业务创新发展。常用的应用场景如下:

精准营销:

基于相关数据进行个性化的商品推荐、视频推荐、旅游商品推荐、广告精准投放等。

关系网络:

分析用户关系网络,进行人群关联,精细化任务画像,扩大营销范围及场景。

库存优化:

分析销售与库存关系,帮助企业智能化进货、退货,合理使用库存控件,并保证健康的供应链关系。

金融风控:

金融领域是算法模型应用比较成熟的领域,应用的场景也很多,比如根据个人信用评估,进行贷款发放的预测,贷款还款的预测,进行风险控制,以及屏蔽羊毛党等营销场景上的使用。

其他:

在不同的行业领域,会有更多的分析场景,如文本分析、图片分类、视频分析等。

截止到今天,袋鼠云企业级一站式数据中台PaaS-数栈具有数据计算引擎、数据开发平台、数据科学平台、数据资产平台、数据服务引擎等5大产品模块,平台覆盖全链路的数据采集、数据分析、数据挖掘、任务运维、数据质量、数据地图、数据模型、数据API开放等场景,充分满足企业建设数据中台过程中的多样复杂需求。

目前,西湖风景名胜区、西溪湿地、山西商务厅、老板电器、中金易云、河南世纪联华、浙江大学、常州旅游商贸职业技术学校、宁波图书馆、京东方、福建票付通、观远数据、东方龙马、佰羚数据等客户和合作伙伴都已经在使用或者参与数栈产品共创。

数栈——企业级—站式数据中台PaaS

数据计算引擎

提升企业数据共享能力,加速释放数据价值

离线计算引擎

基于Apache Spark,计算速度比MapReduce快百倍

实时计算引擎

Based on Apache Flink, high throughput, low latency, high-performance

Data Development Platform

One-stop platform for big data development, rapid and complete data CONSTUCTION

Off-line calculation Development Kit

80% less development time data, the whole data link processing, compatibility open, commercial calculation engine

Real-time computing Development Kit

Financial level flow from the research data processing component 10 times acquisition performance, enhanced source engine Flink

Data synchronization engine FlinkX

Distributed multi-node concurrent read and write, high throughput, support for rich data source

Data Science Platform

Machine learning work platform to explore
visualization experiment to build, integrate a variety of mainstream computing framework and rich algorithmic components
Interactive Notebook develop, explore algorithms provide free environment
offline training model, scheduling and task management support cycle
model online deployment, support online and call model updated version of the model

Data platform assets

Construction of enterprise data center assets, data management standardization

Map data

Visualization asset data center, the data can be understood tube Lifecycle

Data Quality

90% by mass checksum coverage scene, check guarantee consistent bis mass data table row

Data Model

OneData build data models and running quickly clear and understandable data specification

Data service engine

Enhance the enterprise data sharing capabilities, accelerate the release of the data value

data visualization

Let the value visible data, to create a new generation of "four" big screen

Analysis Engine

Ten billion-second query data to achieve multi-dimensional, complex frequency analysis

API data

"Code 0" API generation, multi-access control security to protect data open

Guess you like

Origin yq.aliyun.com/articles/704532