Hello and welcome back to our final screen cast of our course. This is the report out portion where we're bringing all of our information together, kind of putting it together in a final report for our audience. Before I jump into that, I want to just show you just a little bit, for those that are interested in how to come up with themes of your texts, because this will wrap up into the report that I gave. I went through and use scikit-learn and non-negative matrix factorization to actually go through and take a look at the frequency counts of words and terms as we have them across the tweets. Then based on that, go through and attempt to create clusters, right? So some of this is a trial and error. Determining how many groupings that you really have in terms of your cluster and determining how many words you really need to see, from a human's perspective, determine what this topic really is. So we'll look at these a little bit in the report, but just for those that are interested in how you might come across that, here's just a little bit of code that you can take a look at. Okay. All right. Let me jump back into the report itself. So I like to start out just going through and including contents. I kept it in the same project proposal format because I had already started in that and I was thinking of this presentation more for an audience that is looking at a higher level because the hypotheses are kind of at that level too. So we're going to go through and do a review of what questions we are trying to answer, the hypotheses, the approach we're wanting to take. Talk about some of the technical challenges. We'll get into the just the ERD to help people understand how the data was organized. Get into our initial findings. Talk about the deeper analysis and then the hypotheses results in the end. So just to review of what questions we are trying to answer. So you guys have seen these before and in terms of what I was doing in the initial hypotheses and then the approach and so if you were in this section and you varied off of what you were trying to do a bit or maybe you had some other ideas or different things you could include different things in your data analysis approach and talk about those things as you go through and give out your report. When we get into the technical challenges for this analysis, one of the things that I ran into was getting the full text of the tweets early on. So that forced me to kind of step back a little bit, figure out how to go and get more texts, and in this case, required me to go through and do some Twitter scraping but in doing so, I also increased my data set by about four times which was good because then it also lent itself well to doing some thematic analysis later. One of the other limitations I ran into was even though I was using pandasql to convert pandas DataFrames into tables. With it being dependent upon SQLite, some of the SQL was difficult to execute with the things I was trying to do but weren't necessarily supported in SQLite but it was something we could manage drowned. Here, I include the entity relationship diagram as I said I would walk people through, help them understand kind of how the data is connected. Why it's connected the way it is, what we're able to do with that. Then we get into the initial findings. So just like I presented to you in our screen-casts notebooks sessions, this is extracting things from their key relevant points that we'd want to get across. In this case here, we're talking about our initial hypotheses and saying that, "Yes, we were seeing what we expected to see. However there were a few key data points that we wanted to investigate further," and kind of detailing those out. Then I go on to talk about what each of those was and the impacts of them. Then we get into deeper analysis. So we can talk about time. So what's his behavior look like in terms of tweets over time. How about, how often does he tweet during the day and what times does he typically tweet in the evening, in the morning, late morning e.t.c. We can do the same kind of analysis and we did around months of the year. We got into sentiment and polarity analysis. So we can say generally he's positive on the positive side in terms of sentiment analysis. You can actually see here we did an extraction from 2011 and 2019. We didn't include all that data but the relevant point is that he was closer, the central tendency here was closer to neutral as here we're kind of positively shifted. That's really what we're seeing here in the analysis. One of the caveats that you want to consider and you might tell your audiences is that sentiment tools are as good as the libraries that they're built upon and the people that built them. Sometimes, words may be considered in different connotations. They also may be not considered an entity for example here in the case of boring Company. Here we've got definite references to boring Company however, the sentiment analysis picked it up and saying, "Whoa! Okay. Hey, that's not a happy tweet." Also here, we got a connotation issue or is talking about a tunnel boring machine but it picked up on boring and dropped that down as well. Then finally, the deeper analysis on the thematic side. So here's all of our topic clusters that we derived. Then at the bottom here just kind of a human interpretation of what I think each of these clusters is about. So generally, yes it's about Tesla and SpaceX but a lot more detail in terms of functionality and performance and what not. So some interesting things. So if we were to grade ourselves, it's kind of what I'm putting here in our final findings. Yes they were highly connected to Tesla and SpaceX. So this was mostly as expected but we found some additional and other engagements that he does retweet and mentioned SpaceX and Tesla more than others but they're not the only ones that are high on his list. About the topics being related to energy and space. Yes, we kind of take a mostly right for this. But so you can say yes we are talking about solar, he talked about batteries as well space. But there is a lot around performance aspects and benefits of the technology that we maybe didn't really pick up on or didn't anticipate but we did find. Then for the last one as for the tone. It was what we expected that he's tone would be a little more positive. Him being a visionary. Looking out there. What we didn't expect though is that his tone would change over time and maybe that's a result of his overall progress in his fields and the success that his companies have experienced. So with that, good luck to you on your projects and we will talk to you next time.
