 So this is a video, it's a first in a series of forecasting videos. It's going to look a lot like the last plotting time series forecast video I did. But the difference is that this one is going to be connected to a series. This is a checklist of the different things I wanted to cover. So this first video I'm just going to go over plotting. And we'll just hit these as we go. So I have plotting, group statistics, looking at variations, co-variations, make time series predictors. So this is things like moving averages, lags, things like that. The idea is I don't want to do a probabilistic forecast by group. Something new I want to do is I want to look at this ng-boost package and compare it versus the benchmark. The idea in forecasting is that you have your alternative method. This might be something that you consider like a sophisticated model. And you usually compare it with simpler methods, whether that's lags of the past or something like an autoregressive model. So I don't know what that's going to be quite yet, but I just have it marked here as a to-do list. And then also look at metrics for accuracy and metrics for uncertainty. So since this is a probabilistic forecast, we'll be able to look at uncertainty. So the first thing to do is to grab the first link there in the Walmart's M5 dataset. This is on my website. I'll link to it in the description. All I did there was a we get to get that CSV file. And if I use visit data, I can take a look inside. So all I have is three columns, date, estate, ID, and sales. So I chose this dataset. I sliced it and preprocessed it this way so that I can forecast for multiple series, not just one. Now what I'm doing is I'm taking the CSV file and just reading it into df underscore raw. And I just take a little peek at the data. And from here, I'm going to steal from an old script that I wrote. So this is from the previous script I did on time series plotting. Just grabbing chunks of code I think I might want to use. So what I'm doing here is I'm making a date time feature. So I'm just passing the date and making just in case that this date is maybe read as a character. I want to make sure that pandas recognizes it as a date time object, not a string object. And I'm also going to add year on here. Lots of times I like to do group statistics to get a sense of if things are trending up or trending down in EDA. So now I'm just adjusting this chunk of code that is looking at the plot. And I make quite a few adjustments here. So plot 9 is a library that I've been using a little bit more when I'm using Python for plotting. I like it quite a bit because I like ggplot quite a bit. And there's a couple differences. There's not too many differences, but there's a couple differences I still need to look up. Other than that though, I think that plot 9 is a really great library. So here I'm just making sure that my directories are right. When I save a file, it's going to the right spot. I want to automate this though. So every time I save the file, I want to have it pop open. I could use this by just printing the object, but lots of times I'm working in a server, like a SSH onto a server. And so I still like to be in the habit of even if I'm not looking at it directly on my computer, automatically pushing the plot to cloud storage like a S3 bucket or Google Cloud. So what I'm using here is I'm using Enter. Enter is a command line utility that looks for file changes. And once a file changes, it will run a command. It's a really handy tool for automation. If you're curious about Enter, I did a video on using Enter to save files to cloud storage. And I'll link to that in the description. So the plot is getting a little bit closer to what I'm looking for. I'm coloring by each of the states. So California, Texas, and Wisconsin each have their own colors. I don't like the color palette though. I want it to have my branded color palette. I had to look this up really quick just to make sure that I could pass an array to the scale color manual function. So what I'm doing here is I'm just taking a palette from my previous script and I'm going to be passing it to scale color manual. And you'll see the red, the yellow, and the blue. Here I'm playing with different shapes. You might think that this is spending way too much time on plotting. And maybe it is. The thought I have is that looking at... I almost think that you can't look at the data too much. So I'm plotting and I'm seeing the data over and over. Slightly different variations of it. Maybe the colors are different. Maybe the orientation of the different groups are different. So I don't think that it's necessarily lost time to play with your plots a little bit. First of all, it's fun. And second of all, you see the data a lot. So in terms of EDA, just taking time with the data to look at it and play around with how it looks doesn't seem like a waste of time at all for me. So I'm just looking at this checklist that I jotted down. We'll slowly make our way through it. All that was done today was the plotting part. Maybe in future videos we'll be able to hit a little bit more. But I'm excited to keep going and I hope you enjoyed this. Thanks for watching.