Wibidata机器学习平台：让公司具备亚马逊和Google一样的能力

大数据应用提供商Wibidata推出了一个用于搭建实时应用的新平台，让用户更方便获取机器学习能力，电子商务公司也可以提供类似Amazon.com等巨头的体验。

新的WibiEnterprise 3.0平台可以让公司客户用能自我调整的高级分析工具驱动网站，从而随时间推移提供更好的推荐和其他功能，包括更相关搜索结果和定制内容。

Wibidata负责现场运行的副总裁Omer Trajman表示，该平台被设计用来为刚开始使用数据科学的客户服务。“他们并未接受过专业训练，但却有分析背景。他们一直都在做营销分析。机制类似，改变的是数据的提供”。

WibiEnterprise 3.0使用名为Kiji的开源框架搭建，该框架是搭建利用大数据集的应用的通用平台。

Wibidata的核心理念是，考虑到公司客户通常只有几秒钟时间来让消费者参与，而人们使用各种各样的个人设备，只需点击几下就能转投其他竞争者。但有了这点儿数据，通过Wibidata的平台分析消费者的数字互动行为，公司客户就有机会了解消费者的情况。要做到这一点，意味着需要打造一个提供消费者全貌的存储系统。

和Google和亚马逊类似，Wibidata的Kijii框架使用中央存储系统，可以让客户收集所有应用、搜索、订购、喜欢、点击和请求商品信息的用户互动行为。这就是所谓的“以实体为中心的存储系统”，它基本上会存储所有数据，以便让公司用成熟的应用和服务来进行实时查询，并对消费者最近的信息进行操作，以提供定制内容、相关搜索结果和推荐。

Wibidata管理数据的方法与传统数据仓库系统截然不同。在电子商务领域，传统数据仓库系统存储交易信息，如可能进行的购买，或在中心事实表（central fact table）中操纵购物车。对于一家商业零售银行而言，这些数据可能包括账户的信用额度和扣除。库存单位信息或地理位置数据则存储在维度表中，以提供该笔交易的细节信息。

Wibidata表示，这种方法有两个问题。它会变得很贵，且以交易为重心，而不是进行交易的用户。更何况，它在使用历史数据时会变得更复杂，因为必须从不同系统抽取数据，经过清洗后才能与现有交易数据进行整合。

使用WibiEnterprise 3.0的公司包括一家排名前十的零售商，该零售商已经将WibiEnterprise 3.0整合到网站中，以便在网络销售过程中提供相关联的、语境化的购物推荐。还有一家国际商业零售银行也在使用WibiEnterprise 3.0的技术，以综合多个客户数据源，应用债务模型来更好地检测欺诈和信用风险。Opower使用WibiEnterprise 3.0向公共事业提供商客户提供定制报告，解释如何减少能源使用和省钱。世界上最大的SaaS（软件即服务）提供商之一也使用WibiEnterprise 3.0来帮助他们的客户识别潜在消费者。

Wibidata是一个强大的平台，但它也反映了在如此多不同设备中管理数据的复杂性。数据供应无限，但使用这些数据所需的技术需要一个组织增加处理其历史投资的方式。此外，在业务方式变得更以客户而非交易为中心上也有文化障碍。

这对Wibidata这样的新初创公司构成了挑战，也影响了它们。Wibidata等倡导颠覆式的方法，这会让它们直接与甲骨文、Baynote等公司竞争。

无疑，获得像Amazon.com一样的能力变得更加容易了，但随之而来的也有挑战，如消费者的意愿，Wibidata等如何才能在竞争中保持领先等。

英语原文：

Wibidata, a big data application provider, has a new platform for building real-time apps that shows the increasing accessibility of machine learning and how e-commerce companies can provide an experience similar to a giant like Amazon.com.

The new WibiEnterprise 3.0 platform allows a company to power a site with advanced analytics that fine-tunes itself, providing better recommendations and other features over time, including more relevant search results and personalized content.

The platform is designed for the customer who is beginning to use data science, said Omer Trajman, vice president of field operations at Wibidata. “They are not classically trained but they have an analytics background. They have been doing marketing analytics. The mechanics are similar, what has changed is the availability of data.”

WibiEnterprise 3.0 is built on an open source framework called Kiji, which provides a common platform for building applications that leverage large data sets.

At its core, Wibidata is offering a platform that takes into consideration the fact that companies often have just a few seconds to engage their customers. People use all sorts of personal devices and can turn to a competitor with just a few clicks. But with all this data, companies also have an opportunity to learn about their customers by analyzing their digital interactions. Doing that means building a storage system that provides a 360-degree view of the customer.

Like Google and Amazon, Wibidata’s Kijii framework uses a central storage system that allows a customer to collect user interactions across all of its applications, searches, purchases, likes, clicks and requests for product information. It’s what is called an “entity-centric storage system,” which essentially pools all the data so a company with sophisticated apps and services can do real-time queries and act on a customer’s recent information to deliver content personalization, relevant search results and recommendations.

Wibidata’s approach is in contrast to traditional data warehouse systems that manage data in a much different way. In the context of e-commerce, these older systems store transactional information such as likely purchases, or shopping cart manipulations in a central fact table. For a retail bank, this data might include credits and deductions from accounts. SKU information or geographic location data are stored in dimension tables to provide a detailed view of the transaction.

There are two problems with the approach, Wibidata argues. It can get expensive and it centers around the transaction instead of the user that is generating those transactions. Furthermore, it gets even more complex when using historical data, which has to get extracted from other systems, cleansed and then integrated with the current transaction data.

Companies using WibiEnterprise 3.0 include a top 10 retailer which has integrated it with its website to create relevant, contextual shopping recommendations during the online sales process. An international retail bank is also using WibiEnterprise 3.0 technology to combine multiple customer data sources and apply in-house debt models to better detect fraud and credit risks. Opower uses WibiEnterprise 3.0 to deliver personalized reports to utility provider customers explaining how to reduce energy usage and save money. And one of the largest SaaS providers uses WibiEnterprise 3.0 to help their customers identify prospective customers.

Wibidata is a powerful platform but it also reflects the complexity that comes with managing data across so many different devices. There is an infinite data supply but the technology needed to use it means a new way of organization that cuts across the way a company treats its own historical investments. There are the added cultural hurdles that come with a change in business approach that is more customer, than transaction-focused.

These sets of challenges also impact new startups like Wibidata, which are advocating disruptive approaches that put them in direct competition with companies like Oracle and established SaaS providers like Baynote.

There is no doubt that it is getting easier to have the same capabilities as a company like Amazon.com. But the challenges come with the will of the customer and the ability of a company like Wibidata to keep ahead of the competition.

原文：Wibidata Machine Learning Platform Offers Capabilities Comparable To Amazon.com And Google (techcrunch.com)