datacanvas-k8凯发

one-stop service

integrated with visualization and code-based environment, the platform can complete all the procedures for data science applications, ranging from design to production in one stop, which meets the fast-changeable development requirements of data applications.

simplify data preparation

support a variety of data connectors that can be used to access data from various data sources, including local data, external data, and data from databases and data warehouses.

reduce complexity in processing big data

datacanvas allows workflows to automatically be converted to hadoop or spark tasks and submit to execution, and masks complexities of components for big data, which entitles analysts the ability to process big data.

create optimized models in a fast and convenient way

with the container technique and an intuitive drag-and-drop ui, and supportive of multiple language programing, stable and accurate models can be built quickly.

put optimized models into service quickly

data analytics can be done automatically and interactively in batch mode or real time, in local or cluster environment, enormously reducing the time needed.

scalable and reusable model library

the accumulated model library improves efficiency in model development; performance can be extended vertically, cutting costs and enhancing practicability.

real-time

support business in real time and get feedback in millisecond.

high performance

with a high-performance advantage, it can deal with hundreds of millions of messages on a daily basis.

ease of maintenance

complete requirement configuration by using generators, rule engines, parameter configuration, interface operation to get the information needed, lowering costs of system maintenance.

product advantages

supportive of multiple roles

- professional data scientists

develop algorithms and complete advanced analytics in customized working environment.

- data analysts

build models with visualizations, and complete analytics and model building using built-in and customized algorithm modules.

- coders

for those who are familiar with r, python and scala, they can cooperate with each other on this platform, jointly creating business analytic models.

customized algorithms

- open data science

support r, python and scala etc., and allows developers to upload or use external libraries.

- machine learning

integrate multiple machine learning engines and is supportive of teamwork.

support big data analysis

- full amount data processing

access data from hadoop clusters using apache spark and apache impala, instead of only processing sampled data that is compressed.

- high computing load

support calling distributed computing engines such as spark and mapreduce when the analytic flow is running, which can improve efficacy in data processing.

- coders

for those who are familiar with r, python and scala, they can cooperate with each other on this platform, jointly creating business analytic models.

teamwork

- joint development

develop together among team members, improving efficiency in development.

- knowledge sharing

support model sharing, avoiding repeated work.

engineering capability

- version control

track and monitor historical revisions; flexible switch between versions and grey release extremely improve flexibility in data analysis.

- support devops

model developing, debugging, testing, and running in production environment can be completed in one stop, allowing uninterrupted integration and delivery.

automated ops

- schedule and monitor

automated scheduling can be executed as the time set or in a period cycle; globalized monitoring allows you to keep up with the execution of scheduling.

- flexible deployment

the platform can be deployed in a machine room or on cloud, and the cluster size can be adjusted accordingly.

model productionisation

- loading management

model loading management for metric rules, machine learning and deep learning.

- lightning decision

streaming data combined with model computing delivers a fast processing speed, helping to make a lightning fast decision.

- agility

flexible hot configuration, highly reliable design, and offline system docking.

typical application scenarios in financial market

analysis on customers’ footprint of e-banking

analysis on customers’ footprint is the basic condition to predict customers’ consuming behaviors. after analyzing expenditures of customers in different sites, the consuming preference and behavior change of customers can be structured. meanwhile, you can get clues about whether the scope of consuming is widening or narrowing. finally, a dynamic and changeable map of customer consumption is obtained.

big data analysis on customer services

conduct statistical analysis on on-line and off-line customer services according to real-time data to obtain different characteristics within service personnel, and find the factors that affect customer services most, thus to improve service quality based on the results. in addition, the service quality in different branches can be collected to obtain the insight of service discrepancy, and help the branches improve their services.

real-time warning on risks

risk control has always been the key point of financial institutions. therefore, it is quite important to conduct risk management to help enterprises progress smoothly. with real-time analysis, exceptional transactions can be tracked quickly, and relevant staff can take measures more quickly to respond to these transactions. meanwhile, early warning can be sent out more quickly with real time analysis on risks, which greatly prevents risks from being formed.

datacanvas: application of online financial services

project background:

at present, the necessary demand of online shopping platform has increased the provision of products and financial installment service, which not only directly provides customers with more flexible payment methods and significantly improves the user purchase experience, but also increases brand stickiness and attracts potential customers.

solution introduction：

based on the customer portrait and product label system, datacanvas makes full use of big data technology, ai technology and expert strategy to fully enable to mine potential customers and personalized recommendation for different people. through the technology of millisecond data update, user behavior analysis, precision marketing and other technologies, we can improve the product click rate, purchase conversion rate, turnover and other core operating data of the linked bank channel, so as to achieve cost reduction and efficiency increase.

increase of click through rate

increase of conversion rate

cooperative partner

solution function：

portrait and label

according to the basic information of customers, transaction behavior, purchase history, footprint information, channel preference, risk preference, product preference, etc., a complete customer portrait label system is constructed for customer screening and precise product matching.

flexible policy configuration

it flexibly configures the strategy of product side and adjusts the result and order of recommendation.

personalized recommendation

products and services can be accurately pushed to customers through various fields of app, sms, customer center, wechat and other channels.

popular recommendation

according to the historical sales situation, customer preference and channel preference, the hot products are recommended to the target customers.

solution advantage：

real-time computing

with powerful real-time computing power, it can monitor key events, dynamic footprints and recent browsing indicators in real time to capture real-time behaviors and interests, and improve the accuracy of recommendation.

rich algorithm library

embedded classic and cutting-edge recommendation algorithm library, including traditional statistical machine learning and cutting-edge deep learning.

precise matching

according to customer portraits, product portraits and real-time features, accurately match customers with products, channels and fields.

remarkable effect

through rich portrait system and powerful algorithm to recommend products and services in real time, improve the purchase conversion rate.

cooperative partners

beijing zetyun technology co., ltd. (datacanvas) was founded in 2013, focusing on the continuous development and construction of automatic data science platform, focusing on providing a complete set of development platform for data scientists and ai practitioners, and providing comprehensive supporting services for intelligent upgrading and transformation of government and enterprises.

datacanvas is a chinese company independently researched and developed. relying on the domestic and overseas leading data science platform, datacanvas has provided real-time and agile ai capability construction for customers in government, finance, aviation, manufacturing, transportation, education, real estate, internet and other industries. through the datacanvas platform, it can provide automatic machine learning analysis and real-time computing capabilities and help business analysts and data scientists quickly cooperate in development, and realize automatic model creation, management and application support.

to create greater value for customer business on cutting-edge technologies and solutions such as technological innovation and artificial intelligence. the company has outstanding data scientists and product r & d team and has front edge industry practice experience in automatic deep machine learning, data modeling, big data analysis and other fields. in 2020, with the original deeptables open-source project, datacanvas won the first place in the world in the kaggle competition among more than 1100 teams including well-known e-commerce companies and search engine companies. headquartered in beijing and facing the whole country, it has branches in shanghai, shandong and shenzhen.