Hi. Now, let me introduce Seaborn library for plotting. As I already explained, Seaborn is good for drawing statistical graphs. Seaborn is dataset oriented library for plotting on dataframe and array. Especially, Seaborn is tightly linked to its pandas dataframe. In case of drawing graphs based on pandas dataframe, I strongly recommend you to use Seaborn library. Seaborn is especially useful because it internally performs semantic mapping and statistical aggregation to derive or to present informative plots. It is easy to use once you are get familiar with Seaborn. One example. The same logic, canvas and a subplot, only one subplot is created. Figure size is here. In case of Seaborn, there is a set style function. I choose at this time whitegrid. There are many styles provided so you can easily look it up by Googling. Background option is determined by the set on the bar style. You made it to white, dark, ticks, instead of whitegrid. Ax, I'm adding sns.scatterplot. At this time, we are using scatterplot using sepal length and sepal width. Palette is determining color. At this time, I'm choosing bright. There are many options available. You can also Google it. Hue is the color, but color determined by label. At this time, it is iris dataset. Obviously, label, that is the name of variable containing kinds of ideas. Style, label also, depending on which kind. Style is also determined. This is the marker size, style is the marker shape, and dataset is iris. This is the different syntax. In case of sns, using data equal dataset, you are specifying dataset, and then you are using variable names you read in that dataset. You don't have to slide at this time, you simply name variable name. Then it chooses that data up very well. Specific variable. The other command lines all are already explained. Let me show you the outcome. It looks like this. You are easier to draw or generate graph based on pandas dataframe. As you can see here, if you use sepal length, sepal width, many data point is overlapping. In case of setosa, relativity not overlapping. But in case of versicolor and virginica, they are heavily overlapping, so not easy to separate or predict the class of an iris. Then boxplot. At this time, also iris, you are simply specifying data iris. It's so easy. It looks like applying a boxplot function to dataframe. That's what I already explained. Similar technique is used. Sns boxplots simply use call a function and put dataset. Then what happens? At this time, you don't have to specify which variable is numbers and which variable is strings integers. You don't have to specify it. You simply provide data set name, then automatically SNS checks data types, and also if the element is no string and numbers. In that case, it automatically generates box plots or line graphs or scatter plots. That's the benefit of using this SNS. Previously, we need to took out the specific variables containing numbers, but at this time, you don't have to do that. Obviously, this is what we saw before, automatically called laser signed and Fisker. Our total values are presented, also outliers at this time is presented. Obviously, you can add other parliament values in order to change the color and shape. You can do so. Also at this time we are using the same box plot function, but at this time X-axis used for label. What is the label, [PHONETIC]. Those information is included in labeled variable in the iris data set. At this time we are using only sepal width variable. Then what we have is this one. It is so easy to draw figures and you can easily identify patterns. In case of sepal width, the overlapping between first color and [PHONETIC] some overlapping here. Three primes are overlapping in those areas. Joint plan. What is the joint plan? Joint plan provide a histogram to gather scatter plan. Petal width is used. If you look at this shape, you can easily identify or the shape is what we saw. When we draw a scatter plan with petal width, petal length. But if you use joint plan petal l, you specify variables, petal length, petal set, and data-set and at this time KD. KD means that kernel density estimation, kernel density estimation is quite similar to a histogram. In case of histogram, the control of histogram is not smooth. But if you apply KDE, the contour becomes smooth. KDE technique is for smoothing. That's why the contour up here looks like this one and on the right side, petal width contours, density contours, upper side, petal length, density contours. The 60 plus you can get from scatter plan. Obviously, the dense lines means that more data point is [inaudible].It looks like a the map presentation of density. The other case is, as I said before, SNS used for statistical graphics. You can include linear function estimated by the data point. Petal length is used for X variable petal width is used for Y variable, dependent variable, independent variable and you are predicting pedal is based on petal length. There's strong linear relationship and 95 confidence interval is presented here. Implant linear model plaque. As soon as that linear model plot, implot function you are using, you are specifying dataset and you are specifying the petal variables that you are going to use for linear estimation and heighten aspect 1.5. This one actually determines the size of the shape of this figure. What about 1. aspect? Height means that the height and this is the x-axis. Lengths over x-axis, which is 1.6 times over height, that's why we are having this rectangle shape of graph. If you use 1, what do you see? Square shape. Hue, you can add using hue or parameter, that kinds information. Let me show you this one. If you make that parameter active and execute the set again, what you see this one. How nice it is. Aspect is 1.6, then it is better to see the linear relationship. Linear relationship for each kind. The slope is different. In case of setosa and virginica, the slope looks similar but in case of first color, the slope is different. If you add a hue, the whole data, petal length and petal width are subdivided into three groups for setosa, versicolor, virginica, and estimated linear lines for the two variables, and 95 confidence interval is added. How nice it is, wonderful. Violin plots, you can add a violin plot. We are sns.violinplot, and we are specifying dataset names. That four violins, thin violin, thicker violins are presented. What is this one? Reading violin, what is included? A boxplot. Boxplot is included. It means that violin plot shows distribution of data. It is providing more data, more information compared to boxplot. Why? Because boxplot is contained in violin. In case of boxplot, outliers are not shown. In order to see outliers, you need to specifically add a parameter. But violin plot contains or present all the data and reading that violin boxplot is contained. Also, violin has different width. It shows density of observations. Why the remains that more observations is locating in that area? Violin plot is providing richer information than boxplot provide. That's the benefit of using a violin plot. At this time, violins are presented vertically. Surely, you can present violin plot horizontally. There are many options you can choose but I'm not introducing all the options here because it is just introductory course. Finally, we arrived at review question, true or false. Seaborn is especially powerful with Pandas dataframe data for presenting information graphically. True or false? Yes, it is true. That's why you better to use sns, especially if the dataset is Pandas dataframe.
