The huge flows of data enabled by social media, smart phones and the internet have created a revolution in the way we interact with events around the world, especially in the arena of conflict. This environment has given us access to information which can be used for investigative journalism, conflict analysis and the identification of war crimes.

This talk explores the use of open source information in these arenas, specifically by using case studies from Syria and Ukraine.

The answer is a mix of measurements, models and statistics. In this session, explore the observations, predictions and forecast models, and weather data as a variable to consider in machine learning models. Learn how it is done and ways you can use weather and climate data from several examples.

This talk will share a range of examples that covers everything from commuter flows to baboons, cyclists to songbirds in order to demonstrate how maps and data visualisations offer a window into big data. Many of the selected examples started out life in R, you will get the chance to see how R is not just great for data wrangling but for visualisation as well.

Harnessing big data and data science, and integrating them into the heart of company culture is critical to building data-powered businesses. In this talk Mike Hyde will draw on his experiences at Facebook and Skype, and highlight key ingredients and lessons learned on the journey to data-impact when dealing with very large and complex data sets and rapidly moving organisations.

During this talk, you will discover how basic data and AI have the potential to bring the next revolution in operational visibility. You will explore case studies from financial and professional services sectors, where StatusToday’s technology was deployed with direct application to understand employee productivity, identify cyber and human risks (insider threat), and categorise communication patterns for key clients.

In this talk you will focus on the lessons from a recent implementation of a smart search in an instant grocery delivery app.

In Part One you will explore the theory of the Support Vector Machine algorithm which is driving the smart search.

In Part Two, you will learn the agile practices that are needed to put smart search in to production.

This talk will share the intuitions behind this family of algorithms. You will explore some of the Python tools that allow you to implement modern NLP applications, followed with some practical considerations.

1. What Kafka is
2. What the major frameworks are, and how they steer you towards one-by-one message processing
3. Why you shouldn’t do that, including performance measurements for different methods of loading data into a Postgres data warehouse
4. How to avoid one-by-one processing

“Bob’s on vacation – how do I run his model?”

“Is my neural network useless or should I continue tweaking its parameters?”

Have you ever heard any of the above before?

Pawel Subko's team had the same problems when running research and multiple commercial machine/deep learning projects. In this talk you will discover a number of best practices that can significantly improve your team’s performance, based on Pawel's experiences.

You will explore the process of building a robust data science pipeline by using a range of technologies (e.g. Git, Docker or Neptune – Pawel's in-house tool for managing machine learning experiments).

Drawing on his blog post, Sean Owen responds, offering counterpoints from an engineer, in search of a better understanding of how to teach and practice data science in 2017. You will explore some key points in the history of data science from the past 50 years in order to build up a more complete view of how data science sprung out of statistics and merged with computer engineering. Finally, you will discover Donoho’s view of what it means to build data science capability with one taken from the experience organizations doing so in the context of Apache Hadoop, Spark, and other big data tools.

In this talk you will go through the various steps in the algorithm from data preparation and time series decomposition through to finding potentially multiple anomalies. The technique could be used to spot deviations from behavioural patterns with the benefit that it is easy to see why an anomaly is unusual.

See how these detectors can be applied in the context of web media search, advertising and social media, and analyse the precious contribution of computer vision in understanding how people and cultures perceive visual properties, underlining the importance of feature interpretability for this task.

Based on this particular use case, you will learn how to go from a new project to a chatbot handling technical support. You will explore the building blocks of the system like receiving and sending messages, natural language processing and integrations with existing messaging platforms such as Telegram, Messenger or Skype.

The session will cover the following topics:

• The existing chatbot ecosystem
• Performing text analysis on user input
• Identifying the best response to the user
• Personalising the response based on who the user is
• Integrating the chatbot into applications and messaging platforms

During this talk, you will gain knowledge of the components necessary to build chatbot based system, including natural language processing and messages handling. The aim is that viewers will be able to go from never writing a chatbot, to building one which is capable of holding a conversation.

In this highly interactive session, you will discover how to leverage Spark to rapidly mine a large real-world data set. We will conduct the analysis live entirely using an iPython Notebook to show you how easy it can be to get to grips with these technologies. In the first part of the session, we will use a sample of data from the Open Library dataset, and you will learn how to apply common Spark patterns to extract insights and aggregate data. In the second part of the session, you will see how to leverage Spark on Amazon EMR to scale your data processing queries over a cluster of machines and interactively analyse a large data set (100GB) with a Zeppelin Notebook. Along the way you will learn gotchas as well as useful performance and monitoring tips.

You will then learn various use cases Soraya and Roberto have been working on, such as product recommendations, related products, out of stock recommendations, category recommendations and visual browsing. This is followed by an illustration of how various important other functions and elements contribute to the success of a recommender system and what specific challenges they faced in putting their algorithms into a production environment.

The talk will conlude with an outlining of their data science roadmap, which includes context-aware recommendations, session-based recommendations and tensor decomposition techniques.

In this talk, you will learn the chain of events that led up to his capture, his experiences as a hostage, and his eventual release, with many fascinating stories about programming, IT, travel, and survival incorporated along the way.

Recommender systems are paramount for e-business companies. There is an increasing need to take into account all user information to provide the best, most tailored products. One important element is the content that the user actually sees: the visual of the product.

In this talk, you will discover how Dataiku improved an e-business vacation retailer recommender system using the content of images. You'll explore how to leverage open datasets and pre-trained deep learning models to derive user preference information. This transfer learning approach enables companies to use state-of-the-art machine learning methods without having deep learning expertise.

In this talk you will focus on the challenges faced in moving from a communication strategy based around the timings of production and distribution of catalogues towards a more individually personalised strategy fit for a fast-paced online retailer and the techniques and models used to achieve this goal.

Whilst the techniques covered are completely general, the talk concludes with some applications from financial planning and portfolio management. No previous knowledge of mathematical programming is required, but please note that this talk contains formulae (and Python code).

With Oliver Frost, a Data Engineer at Consolidata, you will explore what tools are available to you as a data scientist inside SQL Server 2016 and on Azure. Discover how R adds value where traditional relational databases struggle, how to use ScaleR functions to build predictive models and see how Azure ML can be used to build efficient machine learning pipelines.