This slide displays the screen for coding the LightGBM model. You can find the jupyter notebook file on GitHub, where you can download it and execute it directly. Similar to previous AI models, we first import necessary libraries such as numpy, pandas, and the train test split library. Then, we load the Excel dataset. After assigning X and Y variables, we split the data set into training and testing data, typically using an 80 and 20 ratio. To build a light GBM model we first import the LightGBM library. After importing the library, we convert the data into a data set structure. Then, we put various hyperparameters into a variable called params. Here, multiclass is used for multiclassification, and since the iris data set consists of three species, it is defined as 3. The learning rate is set to 0.1, and the maximum number of leaves for the decision tree is set to 31. Additionally, top_rate represents the proportion of data with the highest gradients in GOSS, and other_rate represents the proportion of data outside the top rate, with the default being 0.1. Then, we train the model using the defined parameters and the training data. As a result, when calculating the accuracy with the training data, we get an accuracy of 96.7% and the same accuracy is observed when applying the testing data. You can create a simple code using the four command to check whether the actual and predicted species match. As shown in this slide, if they match, it will display 'True', if they don't, it will display 'False'. Additionally, you can use the plot uncore importance command to visualize the importance of the features. This shows that the width of the pedal has the greatest impact on classifying the Irish species. Additionally, based on the obtained results, we can calculate precision, recall, and F1 score values. Furthermore, we can also obtain a confusion matrix. Reviewing previous videos on the confusion matrix might be helpful for understanding.