Using Amazon Machine Learning to Predict the Best Time of Day for Exercise – Pt 3: Automating the Model with Alexa

Integration with Alexa allows a user to obtain a workout recommendation (and create a machine learning model) all by voice command.

002

Note: This is the third post about using Amazon Machine Learning to predict workout intensity. Check out Part 1 (Overview) and Part 2 (Building the Machine Learning Model) for background. A working model is available via web and Alexa. Code can be found/downloaded from my Hackster site.

After I was able to build a working model, I needed to come up a way to automate the model. I originally planned to allow access through my website, but decided to use Alexa in addition to the website link.

Note: The process of creating an Alexa skill isn’t too complicated (if you have experience building lambda functions). That being said, I suggest you start by building a sample skill – such as the Fact Skill example. Also, be sure to read and follow the certification requirements.

Alexa, AWS, and the exposed Fitbit APIs provided a mechanism to build a model and return results for a specific user – all initiated by voice.

Step 1 – Linking the user’s Fitbit account to the skill

A user has to link his/her Fitbit account to the skill before s/he can (a) build a specific machine learning model based on their history and (b) get a workout recommendation. Step 1 covers the logic for this functionality.

Click image to enlarge

Slide7

Continue reading “Using Amazon Machine Learning to Predict the Best Time of Day for Exercise – Pt 3: Automating the Model with Alexa”

Using Amazon Machine Learning to Predict the Best Time of Day for Exercise – Pt 2: The Learning Model

Amazon Machine Learning, leveraging activity data from a your Fitbit, can be used to predict workout intensity.

MachineLearning_VideoThumbnail[1]

Note: This is the second post about using Amazon Machine Learning to predict workout intensity. Check out Part 1 (Overview) for background. A third post will cover the Lambda and Alexa code used to automate the model. A working model is available via web and Alexa. Code can be found/downloaded from my Hackster site.

When creating my prediction model, I first had to define workout effectiveness. Was it measured by the total number of minutes that I worked out, or my average heart rate? Should I consider my peak heart rate, or number of calories burned.

After doing some research, I decided that workout intensity would be best captured my the number of “active minutes” for each workout. From the Fitbit website:

All Fitbit trackers calculate active minutes using metabolic equivalents (METs). METs help measure the energy expenditure of various activities. Because they do so in a comparable way among persons of different weights, METs are widely used as indicators for exercise intensity. For example, a MET of 1 indicates a body at rest. Fitbit trackers estimate your MET value in any given minute by calculating the intensity of your activity.

From there, I needed to determine what inputs impacted workout intensity. After more reading, I landed on the following as inputs to a successful workout.

  • Sleep (or lack of sleep)
  • Previous Day’s activities
  • Time of Day
  • Stress
  • Food Intake60700774[1]

My Fitbit tracks sleep and the amount of total activity of a given day (measured with the count of total active calories burned in a day). I can also get the start time of each workout and the resultant active minutes in that workout.

Stress, being subjective, is harder to measure. I originally used the number of meetings in my outlook calendar to determine my level of stress, but found that metric to be inaccurate (at least, without additional research/tuning).

Continue reading “Using Amazon Machine Learning to Predict the Best Time of Day for Exercise – Pt 2: The Learning Model”

Using Amazon Machine Learning to Predict the Best Time of Day for Exercise – Pt 1: Overview

A simple application that uses your Fitbit data to recommend the best time of day to workout.

Best Time of The Day Workout Recommendation

Note: This is the first post about using Amazon Machine Learning to predict workout intensity. Part 2 will cover the design of the Machine Learning Model. The third post will cover the Lambda and Alexa code used to automate the model. A working model is available via web and Alexa. Code can be found/downloaded from my Hackster site.

A few months ago, one of my former co-workers1 published a brilliant post on Linked about using machine learning to predict his mood. It got me thinking – with all the data at my fingertips, what could – or should – I attempt to predict?

I ultimately decided to focus my attention on fitness. Specifically, I wanted to determine the most optimal time to workout on a given day.

Why? Well, when it comes to fitness, we live in a world of paradoxes.

We have the best fitness information available, the most knowledgeable doctors, and affordable devices to track our health.

We have the best workout options: gyms in every neighborhood; gyms at our offices; online and DVD courses to support fitness at home.

We even have more flexibility in our schedules. Many people work from home at least once a week. Our gyms open earlier and close later. Our employers support and encourage an active lifestyle.

Yet despite this, people struggle to find time to work out. And when we do get to the gym, we are either rushed or distracted – so our workouts become less impactful to our overall health.

texting-in-the-gym-300x300[1]

I built my app to help myself, and others, find the most optimal time of the day to workout. Are my workouts more efficient in the morning? Or do I get better results by working out at noon?

Each user is different, so I had to build the app so that each user’s prediction model was customized to their activities. This was done by building a custom model for each individual user of the application.

I’ll talk in detail about how I designed the machine learning model and how the app is architected in parts 2 and 3 of this topic. In the meantime, feel free to use the web based version of the app or the Alexa skill.

Moving from Apache to Nginx Web Server for better WordPress performance

Moving a WordPress site from Apache to Nginx isn’t easy, but is well worth the effort – if you are on a micro instance and memory is at a premium.

kermit_memeI’ve written multiple posts about optimizing my blog to increase speed (here, here, and here)… unfortunately, my web server continues to crash at least once a day (usually due to a spike in traffic).

I started looking to stripping out unnecessary Apache modules, but had too many problems identifying the required modules for WordPress. After 30 minutes of failed server re-starts, I decided to ditch Apache all together and install Nginx as my webserver.

While Apache is the most well know WordPress webserver, Nginx is growing in popularity (WordPress.com uses Nginx on their site). In order to get a clean install, I created a new t2.micro instance, installed Nginx (and php-fpm) and migrated my WordPress files over 1.

There isn’t a lot of documentation on Nginx: I spent most of my time referencing the instructions at the following sites:

The install wasn’t smooth sailing. I had a beast-of-a-time getting my permalinks to work, and I forgot to change a few of my settings in my wordpress.php conf file (which caused my server to constantly crash).

There are still a few things I need to reconfigure – W3 Total Cache works differently in Nginx, so I’ve ignored it. I also have yet to re-install my certificates for https.

All that being said, my website is as fast as it was when I started my blog – probably due to Nginx, as well as the lack of “bloat” on the server (old WordPress plugins and no MySQL install).

My other call-out is that the use of AWS RDS and an Elastic IP made the most complicated parts of the migration simple.

Tuning Apache to Improve t2.micro Performance

A few simple changes to your Apache config file can greatly improve your micro instance performance.

apache-server-logo[1]Last week, I moved my WordPress DB from my t2.micro instance to RDS, with the hope that the change would improve my website performance (I was originally hosting both the web and DB server on the same EC2 instance). The good news is that I no longer get the “database connection” errors I was wont to receive. Bad news is that I’m still having performance problems: now with Apache.

As a stopgap, I configured an alarm to reboot my instance in case I received a status failure. This was a less-than-ideal fix; my server would reboot multiple times a day – sometimes 2-3 times in a 30 minute window!

Last weekend, I finally got a chance to dedicate time to resolving the issue. I spent most of my time looking at the threading options for Apache – prefork, worker, and event. I tried both prefork (the default) and event (which is the less memory-intensive option of the three). I ultimately stayed with prefork, with the following settings (h/t to Marjin van Deele’s post: Tweaking a AWS EC2-micro instance for WordPress)1.

Timeout 30
KeepAlive On
MaxKeepAliveRequests 50
KeepAliveTimeout 10

<IfModule prefork.c>
    StartServers          3
    MinSpareServers       2
    MaxSpareServers       5
    MaxClients            10
    MaxRequestsPerChild   1000
</IfModule>

Two days later, my server seems to be ok. Next step is to strip out some of the preloaded modules from Apache (as outlined in Hayden James’ post: Strip Down Apache to Improve Performance & Memory Efficiency).