You are given the ML Coursework Dataset

$($

access it via the link

$)$

$$to build a multilayer perceptron

model that can be used for forecasting the export value of crop products for a geographical region

three years into the future

$($

e

$.$

g

$.$

$$given historical data up until or at Year

$2$

$9$

$8$

$4$

$,$

$$your model can predict

the export value of crop products for Year

$2$

$9$

$8$

$7$

$)$

$.$

$$Your model is allowed to be either a regression

model or a classification model. For more information about the dataset, see the README file

included in the dataset folder.

What To Submit

You are to submit THREE separate documents clarified in items A

$-$

C below. Do NOT submit them as

zipped

$($

or other compressed

$)$

$$files$.$ Make sure the file name and content for each of the

$3$

$$files

indicate your Candidate No

$.$

$($

and not your name

$)$

$.$

A

$.$

$($

$9$

$0$

$$marks

$)$

$$Report

$-$

$$Prepare a report that answers ALL

$4$

$$questions below. Use the EXACT

question themes highlighted in bold as section titles for your report, so that your report has

$4$

$($

titled

$)$

$$sections exactly. Submit the document as a single pdf file.

$1$

$.$

$($

$1$

$5$

$$marks

$)$

$$Performance

$$

$($

$8$

$$marks max

$)$

$$Report the performance of your model, and also describe

$($

using

maths formulae if appropriate

$)$

$$how the reported performance was computed

from the model outputs.

$$

$($

$7$

$$marks max

$)$

$$State clearly the total number of instances used and then the

number of instances in your training and test sets and how these sets were

derived from the given data.

$2$

$.$

$($

$2$

$0$

$$marks

$)$

$$MLP model

$$

$($

$1$

$1$

$$marks max

$)$

$$Describe your multilayer perceptron

$($

MLP

$)$

$$model$,$ making sure

to specify

$($

including maths formulae when appropriate

$)$

$$at the very least the

activation function for its output layer, the loss function used to train it

$,$

$$and the

number of units in its output layer.

$$

$($

$9$

$$marks max

$)$

$$List and describe the steps that you took to prevent overfitting

for your model

$($

using maths formulae when appropriate

$)$

$.$

$3$

$.$

$($

$3$

$5$

$$marks

$)$

$$Features & Labels

Machine Learning @ University of Sussex

$$

$2$

$0$

$2$

$4$

$($

Temitayo Olugbade

$)$

$$

$($

$1$

$0$

$$marks max

$)$

$$Describe

$($

as a list

$)$

$$how the label

$($

s

$)$

$$was

$/$

were derived from the

given data

$($

using maths formulae when appropriate

$)$

$.$

$$

$($

$2$

$5$

$$marks max

$)$

$$Report the features that you used for your model, making sure

to specify the features used and the total number used, clarify why these

features were selected, and describe

$($

as a list

$)$

$$how they were derived from the

given data, if applicable.

$4$

$.$

$($

$2$

$0$

$$marks

$)$

$$Preprocessing

$$

$($

$2$

$0$

$$marks max

$)$

$$List and describe the preprocessing that you did on your

features for building your model

$($

using maths formulae when appropriate

$)$

$.$

B

$.$

$($

$5$

$$marks

$)$

$$Code

$-$

$$Prepare all the code used to complete all the tasks above and submit as a

single pdf file.

C

$.$

$($

$5$

$$marks

$)$

$$Model outputs for the test set

$($

s

$)$

$-$

$$Prepare the prediction outputs for your model in

a single csv file and using an appropriate header row. The file should have at least

$3$

$$columns:

$1$

$$column for the data instance id

$,$

$1$

$$column for the corresponding true label, at least

$1$

$$column

for the corresponding prediction

$($

s

$)$

$$for your model. Submit the document as a single csv file.