We had off for Thanksgiving

Hi everyone!

This week we talked more about how we were going to get our own data after we scrape the data from Groupon. By getting the deals from Groupon, we’ll create a website that will be a form of a survey. This will be distributed to people so that they can browse the deals we show them, then select which ones they liked. We will then use dimensional reduction techniques to shorten the matrix that will form from the ratings. This will then allow us to “recommend” deals to the participant that we think they would like.

Next week, we have off due to Thanksgiving!

Have a happy holiday!

-Alyssa

Hi everyone!

We’re back! While we had the week off, the team and I still looked into literature review on dimensional reduction techniques for when we start having to analyze our data. We’re looking into two techniques called principal component analysis and single value decomposition.

For next week, our mentor asked us to continue reading up on it and start thinking of ways to use them. We’re going to try to work on examples to get a better understanding on how the techniques work.

Talk to you later!

-Alyssa

This week we had off.

–Alyssa

Hi everyone!

For this week, I just tried to get a working knowledge of GitHub. We’re hoping to be able to use this platform to have access to everyone’s code and, in my case, have help in finding any errors in my code.

Other than learning about GitHub, the team started getting some feedback from companies to start collecting data. So far, we have gotten approved to use the databases from Groupon. While we have put feelers out to other deal sites, since Groupon is such a big site, we’ll have no shortage of data!

That’s all for now! Next week we are having some time off to get a break. So see ya in 2 weeks!

-Alyssa

Hi everyone!

I have even better news than last week! Our mentor helped us figure out how to neaten up the code over the week and it worked! The code outputted what we wanted and did it neatly.

We also set up our group GitHub repository. We all made our own accounts and then joined a repository so that we all have access to the code. It will also help us when it comes to working on it when we’re away from each other.

With this new site, I have a ton of questions and am kind of lost. So for next week, I’m hoping to read up a bit on the site and further understand the workings of it. We are going to have a meeting where we learn about the site so that we can use it as efficiently as possible. So fingers crossed I can learn enough about it that I can have thoughtful questions.

Enjoy the fall weather! See ya next week.

-Alyssa

Hi everyone!

So this week was kind of a hit! Thanks to a partner, the error in my code was found and it’s now working! The problem was that I had a few extra words that were unnecessary and, of course, messing it up. But it’s figured out, and we can move on!

So the result of this fix was that it was outputting the confidence and support values of some of the deals. Which is great, that’s kind of what we want! However, what would be fantastic would be if the code would compute  all  of the deals’ confidence and support values and output only the top 2(or any amount) best values. I thought I already had that written in the code, but I guess the way I had it, it outputted the first 2 values it computed rather than the top 2.

This week is the final week to play around with this code and devote the entirety of my time to understanding it. I’m hoping to complete the code and have a successful output by this Wednesday. For the next few weeks, we are going to start gathering data from deal sites to start working with them.

That’s all for now! Have a nice week!

-Alyssa

Happy October!

So, this week was a continuation of the last week. I worked on figuring out Python, and with the help of my teammates, it’s starting to make sense!

We are using the apyori library which already has the formulas implemented. I know this may sound small to more experienced python users, but I learned how to attach an excel list and how to install a library! Yay!

Something went wrong with one line of code so I’m trying to figure that out in order to move on. However, until then, I will still be working on finishing up this “Alice” example.

Hopefully by next week I would have figured this out and can try another example. Talk to ya then!

-Alyssa

First I apologize for the late posting! I had exams so making a blog post was on the back burner until studying was done.

So! What I did this week was pretty lackluster. I downloaded Python and did my best to learn a basic of it. Since this is a new programming platform to me, it was a serious struggle. Needless to say, I’m still working on understanding it…but it’s coming along!

Our goal for the week was to have a working code that would output a recommendation based on the Alice example. Since I am still in the understanding phase of Python, I didn’t have this and will continue to work on it for this upcoming week.

Since it was mostly a struggle with Python, I wasn’t able to work on much else, so this is a pretty short posting. My goal is to have a little more of an understanding of Python and perhaps have a little more code that would output a recommendation by the next meeting! Fingers crossed!!

-Alyssa

Hi everyone!

So hopefully everything in the last post made sense because this week was a continuation of it. Still using the Alice example, I applied another numerical measure called lift.

Lift tells us exactly how likely one item, like caramel, is bought when another item, like an apple, is also bought. Lift also takes into account the popularity of the first item, so in this case, caramel. The equation looks like:

*In case you’re interested, that equation was done with LaTeX!

Lift helps when it comes to figuring out if there’s a connection or not by:

• If the value is 1, then there is no association;
• If the value is >1, then caramel is likely to be bought if apples are;
• If the value is <1, then caramel is NOT likely to be bought if apples are.

So back to the Alice example, if the lift value was greater than 1, then there was a high chance that she would like the deal.

How can this apply to modeling a recommender system for deals?

This is useful because we could predict that there is a better chance that the next item would be bought, thus the user would want the deal. If the lift value is ≤1, then the model would not offer the deal since there is a non-existent/unlikely association.

For next week, I will be looking into something called “Dimensionality Technique/Reduction”. I’ve never heard of such a thing so this should be pretty interesting. Also, I will be attempting to model a code to create my own “Alice” example. This is also going to be interesting because my background is primarily in math, not coding…it’s going to be a little harder than LaTeX, but I’m looking forward to it!

Stay tuned for more! Have a fantastic week!

–Alyssa