scalable distributed computing framework for real -time analysis...

1
Copyright © 2013 NTT. All Rights Reserved. <Contact>[email protected] Providing real-time analysis for fraud detection, natural disaster forecasting, marketing and making predictions relevant to the stock market, health risks, life sciences, etc. Feature #1: Real-time analysis: High-speed analysis without data accumulation, e.g., complete outlier detection of handwritten images, with 120,000 feature points, within one second. Feature #2: High scalability: Scales linearly with the number of commodity servers (up to 100 servers), e.g., 14.9 times for 16 servers. Feature #3: Profound analysis: Able to deploy sophisticated algorithms such as machine learning algorithms, e.g., Its accuracy rate in handwritten image outlier detection is over 90% in batched processing. S55 Smarter & Instant Analysis of This Huge Surging Present: Jubatus Jubatus is a scalable distributed framework for profound real-time analysis. An accuracy rate of 90% in batch processing was achieved with four times faster (within one second) than batch processing for anomaly detection in handwritten characters. Through our open innovations, we incubate many applications such as “real-time marketing” and “smart management of social infrastructure”. - This technology is collaborative work with Preferred Infrastructure Inc. and with the OSS (Open Source Software) community. http://jubat.us Scalable distributed computing framework for real-time analysis of big data Data processing infrastructure BigData Public & Science Industry Legacy Climate, Security cameras, Medical imaging, … Manufacturing, Customers, Sales, Accounting, ... System Data Logs, Messages, Web analytics, Spam lists, ... Environment Biz/Gov./Social CRM SCM ERP/BI 2 Real-Time Analysis 1 Big Data Stream 3 Smart Action Supported Analysis Engine (Feature #3) Classification : classify data to multi-group Regression : estimate output from input Statistics : estimate entropy of input Recom- mendation : recommend similar data; estimate unknown attributes Graph mining : graph shortest-paths, calculate centrality Anomaly detection : detect outliers from given data NEW Competition - Twitter categorizing - Spam mail judgment - Stock market predictions - Sensor monitoring - Search advertising - Influencer analysis - Traffic analysis etc. … Feature #1 Real-time Batch Simple High-speed Less accurate OSS Sophisticated NTT Group Global Advantage NTT aims at the popularization of Jubatus-related technologies and business, so as to contribute to big data information processing, through world-wide open innovations. Features Application Scenarios Feature #3: Profound

Upload: others

Post on 12-Jul-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Copyright © 2013 NTT. All Rights Reserved.

<Contact>[email protected]

■ Providing real-time analysis for fraud detection, natural disaster forecasting, marketing and making predictions relevant to the stock market, health risks, life sciences, etc.

■ Feature #1: Real-time analysis: High-speed analysis without data accumulation, e.g., complete outlier detection of handwritten images, with 120,000 feature points, within one second.

■ Feature #2: High scalability: Scales linearly with the number of commodity servers (up to 100 servers), e.g., 14.9 times for 16 servers.

■ Feature #3: Profound analysis: Able to deploy sophisticated algorithms such as machine learning algorithms, e.g., Its accuracy rate in handwritten image outlier detection is over 90% in batched processing.

S-55 Smarter & Instant Analysis of This Huge Surging Present: Jubatus

Features

Jubatus is a scalable distributed framework for profound real-time analysis. An accuracy rate of 90% in batch processing was achieved with four times faster (within one second) than batch processing for anomaly detection in handwritten characters. Through our open innovations, we incubate many applications such as “real-time marketing” and “smart management of social infrastructure”.

- This technology is collaborative work with Preferred Infrastructure Inc. and with the OSS (Open Source Software) community. http://jubat.us

今年度から「グローバルアピールポイント」枠を設けます。 3行程度まで。フォントサイズ(12point)、フォント種の変更は不可です。

Scalable distributed computing framework for real-time analysis of big data Data processing

infrastructure

BigData

Public & Science

Industry Legacy

Climate, Security cameras, Medical imaging, …

Manufacturing, Customers, Sales, Accounting, ...

System Data Logs, Messages, Web analytics, Spam lists, ...

Environment Biz/Gov./Social

CRM

SCM

ERP/BI

2 Real-Time Analysis 1 Big Data Stream 3 Smart Action

Supported Analysis Engine (Feature #3) Classification : classify data to multi-group Regression : estimate output from input Statistics : estimate entropy of input Recom- mendation

: recommend similar data; estimate unknown attributes

Graph mining : graph shortest-paths, calculate centrality

Anomaly detection

: detect outliers from given data

NEW

Competition

- Twitter categorizing - Spam mail judgment - Stock market

predictions - Sensor monitoring - Search advertising - Influencer analysis - Traffic analysis etc. …

Feature #1 Real-time Batch

Simple

High-speed Less accurate

OSS Sophisticated NTT Group Global Advantage

NTT aims at the popularization of Jubatus-related technologies and business, so as to contribute to big data information processing, through world-wide open innovations.

Features

Application Scenarios

Feature #3: Profound