In 2019, a crude oil spill struck more than 2,000 kilometers off the coast of Brazil’s Northeast and Southeast, affecting more than 400 beaches in more than 200 different municipalities. The first reports of the spill occurred at the end of August with sightings still spreading later this year. This post explores the sighting records available on the IBAMA website to view the impact of the leak using ggmap and gganimation.

We learn from the masters, doing the math by ourselves, so, in this post we reproduce two data analysis from David Robinson’s #TidyTuesday screencasts, the first one about the horro movie ratings dataset where he uses a lasso regression to predict the ratings of a movie based on genre, cast and plot. The second is about a dataset of pizza ratings in NYC and other cities. There are goods data handling tricks using tidyverse in these analysis.

How good is the Fitbit measures comparing to Polar H7? The wearable Fitbit bracelet measures the heart rate based on the expansion and contraction of the capillaries in the skin throught measurement of the reflection and absorption of LED lights, different from the method used heart rate monitor Polar H7, which captures the electrical signals from the heart beat. In this post, we’ll access a WebAPI using OAuth2.0 to get Fitbit data and compare it with those obtained by a Polar H7, imported from a GPX file during the same training session.

Much of the my work about lean/digital consulting and change management is to make people learning. Learning new concepts, new values, new ways of doing a job or an activity, learning to managing and lead using new tools and methods, and also involves unlearning concepts and points of view built over time, to make room for the new knowledge. So, learning is a key aspect, and also a barrier, for the people involved in the change or transformation processes. In this post, I’ll explore and summarize the key concepts in the great Julie Dirksen’s “Design for How People Learning” book. I frequently review these points when preparing a change roadmap for a team, department or company, the success of a change is direct linked with how people will effectively learn and be capable to use new knowledge in their daily work.

I’ll start series of posts about Keras, a high-level neural networks API developed with a focus on enabling fast experimentation, running on top of TensorFlow, but using its R interface. To start, we’ll review our LeNet implemantation with MXNET for MNIST problem, a traditional “Hello World” in the Neural Network world.

The football season is back, and with it the Fantasy Game! In this post, we will simulate the results and the scoring of my Fantasy League games. To do that, we’ll project the scoring of teams using Monte Carlo simulation with data scraped from sites that predicts players' performances. We will combine the various possible scores of a team’s players to estimate the team’s score distribution, and then compare with the opposing team, and finally compute each team’s chances of winning and losing.