Data Exploration and Visualisation

Hello students, welcome to Kwickprep. A table with a thousand rows tells you almost nothing at a glance. One good chart can tell you the whole story in seconds. Today we learn why we explore data and how to choose a chart. We also learn to spot patterns and outliers, and how to read a simple dashboard.

First, a new term. Data exploration means studying the data carefully to understand it. It is the third stage of the AI project cycle, after data acquisition. We do it before modelling, which means before building the AI model. We use tables, charts and simple numbers like the average to explore.

Why not build the model straight away? Exploring helps us find missing or wrong values, like a blank age or an age of two hundred. It shows trends and patterns, which tell us what the model can learn. It helps us spot outliers, which can mislead the model. And it helps us choose the right kind of model for the problem.

Remember a famous saying, garbage in, garbage out. If the data is bad, the model will also be bad, however clever it is. A model cannot fix wrong data by itself. So we clean and understand the data first, and only then we build.

Every chart answers a different question, so start with the question. To compare groups, use a bar chart, like marks in each subject. To see change over time, use a line chart, like the temperature each day. To show parts of a whole, use a pie chart, like how a monthly budget is spent. To see if two values are linked, use a scatter plot, like study hours against marks. To see how values are spread, use a histogram, like the heights of students in a class.

Pause and predict the best chart for each question. Runs in each over, in order, change over the match, so a line chart works, and a bar chart for each over also works. The share of votes is parts of a whole, so a pie chart fits. Height and weight are two values that may be linked, so use a scatter plot.

Now let us explore a real looking dataset of samosas sold in a school canteen. On Monday, forty samosas were sold. On Tuesday, forty two. On Wednesday, forty five. On Thursday, forty seven. On Friday, suddenly one hundred and twenty. On Saturday, fifty.

What do you notice? A pattern is something that repeats or moves in one direction, called a trend. Leaving out Friday for now, sales rise a little every day, from forty to fifty. When we draw this as a line chart, the gentle upward slope is easy to see.

Now look at Friday. An outlier is a value that is far away from all the other values. Friday's one hundred and twenty is clearly an outlier. Before doing anything, find the reason, was it a typing error, or a real event? Suppose there was a sports day on Friday, so it is real, and we keep it with a note. If it were a typing error for forty eight, we would fix it.

Why does one value matter so much? With Friday included, the six days add up to three hundred and forty four, so the average is about fifty seven. Without Friday, the other five days average about forty five. So one outlier pulled the average up by about twelve samosas, and the canteen would cook too many every day.

Finally, a dashboard. A dashboard is one screen that shows several charts and key numbers together, like the dashboard of a car. It answers the most important questions at a glance. You can make simple dashboards in tools like Google Sheets or Excel, and companies also use tools like Power BI.

Let us walk through a simple canteen dashboard, panel by panel. At the top, a big number shows three hundred and forty four samosas sold this week. A line chart shows sales by day, and answers, are sales rising? A bar chart compares items like samosa, tea and sandwich, and answers, what sells best? A pie chart shows cash versus UPI payments, and answers, how do students pay?

Here is how to read any dashboard well. First, read the title and the units, like rupees or pieces. Next, check that a bar chart axis starts at zero, because a cut axis makes small differences look huge. Then look for the overall trend, and after that, the outliers.

Let us revise what we learned today. We explore data before modelling, because garbage in means garbage out. A bar chart compares groups, a line chart shows change over time, and a pie chart shows parts of a whole. A scatter plot shows if two values are linked, and a histogram shows spread. A pattern repeats or trends, while an outlier stands far away from the rest. And a dashboard puts the key charts on one screen.