top of page
Machine Learning Projects
Problem Statement 1:
You are required to implement the following tasks as part of your project:
1. Implement linear regression and find the best possible concept to predict
the probability on AFFA-1
2. Implement case-based reasoning and find the best possible model to pre-
dict the probability on AFFA-1
3. Implement case-based reasoning and find the best possible model to pre-
dict the most appropriate configuration script / class on AFFA-2
4. Implement ID-3 decision tree algorithm (from scratch without using any
library) and find the best possible model to predict the most appropriate
configuration script / class on AFFA-2
Each task involves an empirical study to find out the best possible model for
the given scenario. You are required to use the appropriate evaluation measures
and validation methods to conclude your findings. For some of the tasks, you
may have to pre-process the data appropriately.
Project report:
Problem Statement 2:
Recall your implementation of case-based reasoning in Project-1:
1. When a new problem has to be solved, it needs to be compared with all
of the existing cases.
2. If we have to deploy this solution for a practical problem, it has ‘retain’
phase as well. This leads to continuous increment in number of cases to
be compared.
3. Your task is to overcome this efficiency bottleneck.
4. The algorithmic hint is to cluster your case-base into appropriate number
of clusters using a clustering algorithm. When a new case comes, it should
be classified in the most relevant cluster. Nearest neighborhood should be
computed from the identified cluster only.
You are required to try at least two different clustering algorithms. You may
reuse your code of Project-1. Your main task to study and present a deep under-
standing of performance of CBR in the clustered framework. Your study must
report detailed empirical results on how the performance is effected (positively
or negatively) in this approach. You should also present a detailed study on the
efficiency achieved in terms of number of comparisons in different situations.
Project report 2:
Problem Statement 3:
Your task is to extend MNIST standard implementation to develop following features: 1. Gather your own data of hand-written digits (at least 200 unique sample per project) 2. Test the MNIST implementation on your data 3. Your program should be able to convert an image containing an integer of up to 5 digits into equivalent number. For example, 45321, 88881, 23019 etc 4. You are free to jump to CNN implementation if required
Project report 3:
bottom of page