You are expected to visit data repositories such as Kaggle, UCI

Machine learning repository, PROMISE, etc. and extract three

different datasets in the areas of classification, clustering and

regression respectively. Note that the dataset for classification

should be unbalanced $[3$ marks$].$

$$ Preprocess the datasets based on the following issues to

address: $[5$ marks$]$

o Missing data values

o Duplicate instances

o Outlier detection

o Influential datapoint detection

o Checking normality of the set of features

o Data transformation

o Feature selection

$$ In the case of the unbalanced dataset, consider an appropriate

approach to balance the dataset. Eg$.$ You can use an

oversampling technique such as SMOTE, ADASYN, MAHAKIL,

etc to balance the dataset $[6$ marks$].$

$$ Select appropriate learners $($at least $2$ learners for each dataset$)$

for the training and validation needs and justify from literature

why those learners are relevant for such datasets $[6$ marks$].$

$$ Consider the following approaches for training and validating

the models:

o K$-$fold cross validation $[3$ marks$]$

o Leave$-$one$-$out cross validation $[3$ marks$]$

o Percentage split, E$.$g$.$: $70\%$ for training and $30\%$ for

validation $[3$ marks$]$

$$ Consider appropriate evaluation measures $($across the

approaches for training and validation$)$ to assess the

performance of the models. Select the best model$/$learner based

on a good justification $[6$ marks$].$

$$ Per results obtained from the prediction, classification and

clustering across the three datasets, provide necessary plots to

aid in visualizing the results obtained $[6$ marks$].$

$$ Use a set of hold$-$out data to predict, classify or cluster into the

right bins and provide a good visuals$/$plots of the results $[4$

marks$].$

$$ Discuss your visualized results $[5$ marks$]$