The Predictive APIs and Apps conference — PAPIs.io — says it is the first of its kind, aimed at giving voice to the increasing number of API products offering predictive analytics services. ProgrammableWeb spoke with Louis Dorard, general chair of the conference and author of Bootstrapping Machine Learning, about what attendees can expect to gain from participating.
The First International Conference on Predictive APIs and Apps will be held in Barcelona, Spain, Nov. 17-18. The event has been timed to draw in participants who may be attending O’Reilly’s Strata Conference being held immediately afterward, but is open to any data scientists, developers, hackers and businesses “who are looking to join the predictive revolution and to get value from data,” Dorard says.
Dorard draws attention to the event program, published on Lanyrd:
Several companies who are building predictive APIs and tools to make predictive app development easier will be at PAPIs (BigML, Datagami, Dataiku, Indico, Intuitics, GraphLab, Openscoring, PredictionIO, RapidMiner, Yhat). We're expecting to see both actual and potential users who will share and learn how to use these products. Newcomers will learn and get inspiration from the keynotes, showcases and practical "predictive for all" user stories. Experts will also be interested in the sessions on technical challenges and in the panel discussion on the future of predictive APIs.
Dorard says the initial plan was to create a space for early entrants in predictive analytics to come together, but this year’s quick ramp-up of predictive API products demonstrated the need for an industry-wide, international conference for the fledgling tech subsector. Says Dorard:
Francisco Martin (PAPIs.io '14 program chair and CEO of BigML) and I first talked about setting up such a conference about a year ago. Our initial idea was to bring together people who shared our vision of making predictive technology accessible, to talk about interoperability of tools and APIs, to discuss their future, and to announce new features and products.
As someone who educates people around predictive APIs, one aspect I thought was important was to have people who are interested in the field see what others are building, what challenges they're having and what value they're getting from predictive.
By August this year, we decided to go ahead and create PAPIs.io. We’ve seen significant growth this year in the number of APIs and tools available to make predictive app development easier. It's actually getting harder to navigate this space; some predictive APIs require some knowledge of machine learning algorithms, others not, some make it possible to create your own predictive model from your own data, others expose fixed models. ... Getting the makers of these APIs together can help structure our collective industry experience.
Predictive APIs and machine-learning-as-a-service are also getting a lot of attention since big players like Microsoft and Wolfram joined. They also understand that there's a talent shortage that prevents some customers from applying machine learning, and to get value from their data, and they are working at removing barriers to entry.
Forrester Research last year recognized predictive technologies as “the next thing in app development." This year has definitely seen strong signs of the ramping up of predictive analytics and machine learning as a significant API subsector. Early market entrants like SwiftIQ — which has built powerful predictive and big data analytics tools — have reached a growth stage this year that is requiring them to specialize in a particular niche in order to manage their growth potential (for now, SwiftIQ is focused on the retail sector, for example). Meanwhile, API service provider Apigee has doubled down on predictive analytics as central to its value proposition: its industry leadership and user conference focused on how predictive analytics should feed into overall API design and development in order to help businesses create “adaptive apps.” The 2014 summer conference season was filled with announcements of predictive tools available via APIs: MindMeld, MonkeyLearn, Lingo24, GoodData and FirstRain all opened up machine learning tools or expanded predictive features in their products.
One area where predictive APIs are beginning to make a difference in the adaptive apps space is in heightening the potential to create a killer user experience in the hyperlocal domain. To date, personalization and context have not been sufficient in themselves to create a truly must-have app user experience. Developers are still stymied when trying to create apps that draw on an individual user's preferences, his location and time, and the world around him in order to surface the right type of information and create that essential experience that any mobile user or wearable device owner can't do without, now that he has it.
Dorard believes this is about to change thanks to predictive APIs:
We're all carrying sensors in our pockets (smartphones, smart watches, bands) that provide data on what we're doing at any point of the day. Analyzing this data and enhancing it with any available data that's relevant — open or not, in relation with our geographical position, time, weather, etc. — can create a rich representation of the "context" we're in. We can also measure the actions we perform and record when we perform them. Context representations are getting rich enough to motivate looking for relationships between them and subsequent actions. We can then use predictive analytics to create models that map this context to action and thus to predict or to recommend an action at any point in time based on the context a user is in.
This is the vision behind Snips Pocket Brain, for instance. It's an app (in private beta) that learns from you and predicts what you're going to need when you open it: get transport information, a recommendation for a place to have lunch or to send a text to the person you're meeting next.
Day two of PAPIs.io will showcase a broader selection of use cases where predictive analytics can disrupt other industries. For example, as the smart cities agenda grows, it is expected that predictive APIs will become an increasingly used tool to help cities service their growing populations while managing resources efficiently. Dorard explains:
Better resource management is one of the top value propositions of predictive analytics and something we're seeing in our innovation showcase sessions. Some of them tackle urban challenges. They're very innovative but somewhat experimental at this point.
One of the challenges we're facing is population growth in cities that don't scale well. We're often stuck using the same resources, so we have to be smarter in the way we use them. The usual problem is having too many people who want to access the same resource at the same time. This can be transport infrastructures or recycling centers, for instance. We have to find ways to spread the load. One way to do that is to have people anticipate and adapt their actions based on the current state of the resource they want to use and on predictions of future states. These predictions are usually made available through apps that make use of predictive APIs (this is the case of Qucit, for instance, which is using predictive APIs to manage bike-share access). They can also be used by maintenance teams to increase resource availability (as is the case in the public bikes example).
Interested developers can register online to attend.