Whenever you spot a trend plotted against time, you would be looking at a time series. The de facto choice for studying financial market performance and weather forecasts, time series are one of the most pervasive analysis techniques because of its inextricable relation to time - we are always interested to foretell the future.
In the 2nd part of our tutorial on artificial neural networks, we cover 3 techniques to improve prediction accuracy: distortion, mini-batch gradient descent and dropout.
Do you know what gives red and white wine their colors? Use k-NN to discover the chemical make-up that defines typical types of wines, as well as to detect atypical ones.
Learn how random forests, an ensemble of decision trees, can help predict where and when a crime will happen in San Francisco, California.
Decision trees can be used to identify customer profiles or to predict who will resign. Using the Titanic dataset, learn about its advantages and pitfalls, as well as better alternatives.
You are exploring the nutritional content of food. How can food items be differentiated? How might they be classified? PCA derives underlying variables that help you slice your data for these insights.
Want to use deep learning for your analysis but don't know where to start? This tutorial teaches you how to build your own deep learning box, from hardware purchase to software installation.
Using weapons trade data, we map out who's against who in the complex arena of international politics.
You own a store. How do you discover purchasing patterns, such as which items tend to be bought together? Knowing this can improve your product placement and advertisement.
Modern smartphone apps allow you to recognize handwriting and convert them into typed words. We look at how we can train our own neural network algorithm to do this.
You have employees. But who should you pick to lead them? Learn how to predict leadership potential using multiple sources of personnel data, as well as pitfalls to watch out for.
While an artificial neural network could learn to recognize a cat on the left, it would not recognize the same cat if it appeared on the right. To solve this problem, we introduce convolutional neural networks.
You want to publish ads for your product. While you have 2 promising ad designs, you have a limited budget. How can you find out which ad is more effective, while maximizing the impact of all the ads you publish?
You have customers. But how should you categorize them to target sales? How many of such categories exist? To answer these questions, we can use cluster analysis.
In the 2015 hackathon organized by Singapore's Ministry of Defense, one of the tasks was to predict resignation rates in the military, using anonymized data on 23,000 personnel. Our team won overall 3rd place. In this post, we elaborate on our methodology.
Outliers can be detected by algorithms used for predictions. To illustrate, we use the k-nearest neighbor (kNN) clustering algorithm.
Latent Dirichlet allocation (LDA) is a technique that automatically discovers topics that a set of documents contain. It is used to analyze large volumes of text efficiently. To find out how it works, check out this tutorial.
Singapore turns 50 years old in 2015. While Singaporeans are proud of our progress from 3rd to 1st world status - one wonders how this progress has been portrayed though the lens of global media. By examining Singapore-related news, could we predict Singapore's growth trends? Could we examine how much an export-dependent economy like Singapore is affected by world events?
Research has shown that we like people similar to ourselves. But does this rule of attraction apply to our liking for fictional characters? Analysis of Star Wars character fans suggests so. Personality scores of Facebook users who had 'liked' Star Wars character pages were aggregated and profiled in this post.