This is a solitary task that centers around the utilization of Artificial Neural Networks $($ANN$).$ Your goal is to optimize the model and suggest the most architecturally performant model.

It is necessary for you to:

$1.$ Look through a dataset of your choosing that has at least $10$ variables, including the outcome, and more than $500$ observations. While concentrating on a classification work is advised, a regression problem is also a good choice.

$2.$ Use the dataset to create an ANN from scratch after pre$-$processing, selecting the hyperparameters with common sense. By adjusting the hyperparameters, you can make the model better. Then, using the right metrics, you can select the best model from your experiments.

$3.$ Second, you must implement an additional version of your ANN, this time utilizing the Python sklearn MLP $($Multi$-$Layer Perceptron$)$ library. As a result, you can compare the outcomes obtained in $(2)$ using the same number of layers, neurons, activation functions, etc.

$4.$ Report: The dataset must be fully described in your report, along with the methods and setups you used for the various tests. Don't forget to discuss the hardware architecture that the tests were run on$,$ and in the end, give a thorough explanation utilizing a variety of evaluation matrices, including accuracy, sensitivity$,$ F$1$ score, AUC, etc. You must describe your evaluation procedures and the choices you made to prevent incidents of overfitting or underfitting. The following sections ought to be included in the report: Abstract, Introduction, Methodology, Results, Restrictions, Conclusion, and References.