Showing posts with label Analytics. Show all posts
Showing posts with label Analytics. Show all posts

Wednesday, November 16, 2016

Advanced Analytics Process

Through my time in analytics and through studying different courses I've been following the process laid out in the Coursera JHU Data Science Specialization Track. It looks like this:

  1. Define the question
  2. Define the ideal data set
  3. Determine what data you can access
  4. Obtain the data
  5. Clean the data
  6. Exploratory data analysis
  7. Statistical prediction/modeling Interpret results
  8. Challenge results
  9. Synthesize/write up results
  10. Create reproducible code
  11. Distribute results to other people
Without a doubt it helps you reach the best conclusions when trying to solve large analytical questions.


Thursday, August 18, 2016

Exploratory Data Analysis


I finished Coursera Johns Hopkins University Exploratory Data Analysis course earlier this year. It was a great course for getting into data and using R to get a good idea of what you are looking at. Here are some of the main subjects of the course and few quick thoughts on each.

Exploratory Graphs - You can use these to do a quick and dirty look into the data that you have and see what it might tell you. Then you can put together a more polished set of charts after you gain direction.

Plotting Systems - Here are the ones that are covered:

  1. base - The basic plotting system. You can create plots and then annotate and add things as you go. Easy to use, but set of commands to recreate.
  2. lattice - Everything is created in one function call and no way to add anything after that.
  3. ggplot2 - A combination of the two. A lot of the basics taken care of up front intuitively and then can add if needed after creation.

Clustering

  1. Hierarchical or K Means - Use distances (euclidian or manhattan) and dendrograms to create clusters of data of which to analyze.
  2. Dimension Reduction - Use principal component analysis (PCA) or singular value decomposition (SVD) to trim down the data and find meaningful relationships.

Thursday, July 7, 2016

Federal Reserve Economic Data & R

I was looking for data sets one day and I found a very cool site of tons of sets. The site is the Federal Reserve Economic Data.

I downloaded some Michigan population data by year and put it into R to create a plot with a linear regression.

Here's my code:
mipop = read.csv("mipop.csv")
attach(mipop)
model = lm(pop ~ year)
plot(year,pop)
abline(model)

And my plot:



It was just a fun little project to learn a few new basic skills, but pretty fun.



Wednesday, June 8, 2016

Machine Learning Algorithm Cheat Sheet



As I get farther into data and analytics I've found that having a little cheat sheets to refer back to can be really helpful. That said I've found one that has a ton of information on it, but is still clean and easy to use.

Check it out: https://azure.microsoft.com/en-us/documentation/articles/machine-learning-algorithm-cheat-sheet/

Wednesday, May 11, 2016

Data Mining Resources


I recently ran across a great resource for learning all kinds of things about data mining, analytics and programming around big data.

My favorites are the articles around learning SQL and R. I would also focus in on some of the other online resources. The books aren't as appealing to me and therefore I didn't try any of them, but if you did let me know in the comments.

Check it out here: http://blog.import.io/post/38-great-resources-for-learning-data-mining-concepts-and-techniques.

Wednesday, October 21, 2015

Census Explorer


When you want to generate reports or find trends you have to have lots of data. That means tons of numbers and other identifiers, usually, in a spreadsheet or database. However, that doesn't mean you have to look at it that way. And usually that's not the best way. Taking that data and presenting it in a visual fashion can be more powerful then just the numbers themselves.

The Census Explorer is a great example of that. Check it out here:
http://www.census.gov/censusexplorer/censusexplorer.html.

Wednesday, December 17, 2014

Tableau on your Blog

Sadly, it's been one of those years that flies by. Back in January I saw a tweet about how to get Tableau on your blog in under 3 minutes. I haven't been able to give myself 3 minutes to do that. That said I wanted to share it here so I would remember to do it in 2015. Check out the video at Tableau's blog.

Thursday, November 20, 2014

The State of the World's Children 2014 In Numbers


Back in February I saw the following tweet.
The numbers and information in the report are pretty staggering. The sheer amount of data collected to do all of this analysis is pretty impressive. I suggest reading through the report and understanding the data and the process.

Read more (http://www.unicef.org/sowc2014/numbers/)

Tuesday, June 3, 2014

Coursera Data Science Specialization


One of the items that has me the most excited in my professional development, currently, is the Coursera Johns Hopkins University Data Science Specialization. Coursera touts these specializations as "Your Pathway to Expertise." I guess we'll have to see how true that is, but to this point it has been a great tool for learning data science pretty comprehensively. Essentially, there is a set of courses offered that follow the steps to learn a topic.

For the data science specialization there are 9 courses and the capstone project. I've only taken the first three courses, but here are some of the positives I've experienced about the program so far.

Program Speed and Starting Point

The program overall starts a bit slow, but I think that's almost preferable. Some of the projects and even quizzes can take a bit of time if you are new to GitHub, R or even Coursera. However, the first two courses are referenced in the other following courses so I think it's good to get a good base there. I took the first two at the same time and I feel like that kept me very busy, but helped provide a fuller picture of the specialization.

Course Order

After the first two courses you can take the courses in almost any order. This could be really helpful if you wanted to pick and choose which courses to take. I plan to take them all so it seems easy to take them in the order they are set up.

Redundancy & Length

To this point most of the courses are being offered once a month and they last for four weeks. In my opinion, this has been a great setup. It keeps you fresh and focused and offers you the chance to take something a month later if you need to retake it or just want to take a month off.

Lectures & Quizzes

The lectures and quizzes are all available at the beginning of the course and the lectures are available for download in various formats. For someone who likes to work ahead this has been a huge selling point for me. It just allows you to work at your own pace, as long as your pace is ahead of the weekly quiz deadlines! For the busy professional this can be a life saver.

Projects

The projects have been the hardest part of the courses for me. I feel like I've probably learned the most, but most of that has been because of the discussion forums. The directions and grading clarity has been a big rough around the edges. I believe that's mostly due to this being the first pass through the courses. I'm hopefully it will get better as time goes on.

Overall

If you want to learn more about data science I would definitely check out these offerings.


Read more: Coursera Johns Hopkins University Data Science Specialization

Thursday, May 22, 2014

The 13 Best Maps of 2013

An item that's been on my list to learn more about for a while (and that I vocalized recently) has been visualizing data specifically geographically. Geovisualization (which I'm using loosely) is intriguing because it is engaging even for the non-data nerds. I think because of that mass appeal it can have a huge impact when used correctly.

Recently, I stumbled across this collection of the best maps from 2013. Here are my favorites, in no particular order, a few notes about each:
  • Popular Baby Names - This set of maps is easily consumed at the state level. Any deeper level of analysis would make it too hard to process. I like how they have six full decades of male and female names. It's cool to watch how names come into popularity.
  • Craft Beer - This one doesn't have as much data. And by that I guess I mean it doesn't feel like it has as much data. Regardless, it's about beer. Beer, people. I really enjoyed the fastest growing breweries and also being able to hover each one to get a bit more information.
Closest Pizza Places

  • Closest Pizza Place - While this one isn't interactive I liked how it was broken down to the county level. That's a lot of detail about pizza place location.
  • Adjusted Country Size based on Population - This was one of those that didn't really shock me, but it's not something I think about everyday and then when you see it visualized it really helps you remember how the world's population is distributed.

Tuesday, March 18, 2014

Learning with DataCamp

On February 24 I signed up for DataCamp (datacamp.com) to learn R and learn more about data analysis. I use data analysis regularly, but I had a feeling that using it this way would be a bit different and expand my horizons.

I've completed the introductory course (Introduction to R). Here are are my initial reactions:
  • It's weird to set variables using <-. The equal sign is just so intuitive and used in all other cases I can recall. Although I've just started the second course and it looks like you can set variables with the equal (=) sign as well.
  • I don't ever remember using things like vectors in previous programming experiences. Maybe that was because I've been focused on having a program do something rather than manipulate data that is fed in.
  • Vectors are awesome.
Here's a screen shot of one of the lessons in the first course.



I'm hoping that the rest of the courses will be a great learning experience. The interface is very similar to Codecademy (codecademy.com). Both seem like a great way to learn on your own schedule.

Tuesday, March 11, 2014

Learning in 2014

I've hat a few items on my list to learn for a while. In no particular order:
  • Biostatistics
  • Sports Analytics
  • Visualizing Data (especially geographically)
  • R
  • Linux (and run in on a machine with regular use)
  • Python (started this in 2013)
  • Tableau
  • Master Data Management
I don't know that I will be able to tackle them all in 2014, but I'd like to start down the path to learning a couple of them in a bit more depth. With that in mind I've taken two steps in the right direction.

The first is learning R. I signed up for DataCamp (www.datacamp.com). I've recently started their Introduction to R course and have completed the first chapter (Intro to Basics). I'll report back on my experience as I complete further chapters. Inside of DataCamp there is also a few additional courses on basic data analysis which I plan on taking as well to brush up or expand my skills.

The second thing I've done is signed up for The Data Scientist's Toolbox on Coursera (through John Hopkins University). This is part of the Data Science Specialization offered on Coursera. I'm excited to start that on April 7. I will be doing a little blogging on that as well.