Normalization

3 X norm ¿ 6.0 − 4.7 6.7 − 4.7 = 0.65 X norm ¿ 6.7 − 4.7 6.7 − 4.7 = 1 X norm ¿ 5.6 − 4.7 6.7 − 4.7 = 0.45 X norm ¿ 5.8 − 4.7 6.7 − 4.7 = 0.55 X norm ¿ 4.7 − 4.7 6.7 − 4.7 = 0 X norm ¿ 5.4 − 4.7 6.7 − 4.7 = 0.35 X norm ¿ 5.5 − 4.7 6.7 − 4.7 = 0.4 X norm ¿ 6.3 − 4.7 6.7 − 4.7 =0.8 X norm ¿ 6.2 − 4.7 6.7 − 4.7 = 0.75 X norm ¿ 5.9 − 4.7 6.7 − 4.7 = 0.6 Calculating X norm for each data point in feature 2 X norm ¿ x − 2.3 3.5 − 2.3 X norm ¿ 3.5 − 2.3 3.5 − 2.3 = 1 X norm ¿ 3.0 − 2.3 3.5 − 2.3 = 0.583 X norm ¿ 3.0 − 2.3 3.5 − 2.3 = 0.583 X norm ¿ 3.3 − 2.3 3.5 − 2.3 = 0.833 X norm ¿ 2.8 − 2.3 3.5 − 2.3 = 0.417

6 X norm ¿ 3.0 − 3.0167 0.2749 = -0.6086 X norm ¿ 2.3 − 3.0167 0.2749 = -2.9102 X norm ¿ 2.9 − 3.0167 0.2749 = -0.8379 X norm ¿ 2.8 − 3.0167 0.2749 = -1.2234 X norm ¿ 3.0 − 3.0167 0.2749 = -0.6086 Visualization , visualize it using a scatter plot. Plot the points for each class with different colours or shape to observe the distribution of the data. Manual Implementation (New Instances) Normalization of New Instances | Feature 1 | Feature 2 | | 5.0 | 3.2 | | 6.1 | 2.8 |

9 Recall= T P TP + FN = 470 470 + 34 = 0.932540 Precision (Positive Predictive Value): Precision= = T P TP + FP = 470 470 + 46 = 0.9109 Accuracy: Accuracy= T P + TN TP + TN + FP + FN = 470 + 450 470 + 450 + 46 + 34 = 920/1000 = 0.92 F1 Score: F1 Score= Precision ×Recall Precision + Recall = 0.9109 ∗ 0.932540 0.9109 + 0.93254 = 0.46079649242 Conclusion One of the most basic methods in machine learning is the K-nearest neighbors (KNN) algorithm which has proven to be extremely useful for both classification and regression purposes. It belongs to the lazy learning group, since it adopts a “lazy” method in saving training examples and defers calculation till prediction is required. Unlike eagerness learning algorithms for instance decision tree and neural networks that build a model during the process of training, this technique avoids training models in its entirety altogether (Henderi et al., 2021). In essence, KNN rests upon similarity measures of mainly the distances to categorize novel cases. KNN finds out the k closest neighbors to one of the previously unseen datapoints that is picked according to some distance measure when asking for the prediction. These neighbors' class labels influence the classification of the new instance: it is labeling a new data point by belonging to the most popular among the k neighbors. The choice of a distance metric determines largely the algorithm performance. Among common distance measures are Euclidean, Manhattan, Minkowski, etc. These influence what kind of evaluation of distance between different points of data is made by KNN. This could be illustrated by the example that the Euclidian distance determines the straight line distance between any two points located in space and the Manhattan one finds the sum of absolute differences, which have to do with the same characteristics ( Pandey & Jain, 2017) . The understanding of KNN also includes knowing of some important parameters that include k, denoting the number in this case of neighbors being used in classifying the model. The selection of 'k' influences the model's bias- variance tradeoff: it is possible that smaller ‘k’ values could result into including noise present in the data as part of the model; conversely, larger ‘k’ values might average over larger areas and thus over-simplify the model. Given a specific Iris dataset (i.e., consisting of twelve records, each having two features and three different classes), the computational operation here will be done by KNN which will