DS 807 - Modeling Unstructured Data - Spring 2023 Term 2
Using the IMDB Movie Dataset to predict gross revenue and identify trends.
Project Details:
You can pick any type of data that satisfies at least one of the following criteria:
- Text Data
- Image Data
- Unsupervised Data
Note
Since there is a great range of potential data, the instructions are written in a general way. If some steps do not make sense in your case, please reach out to verify.
Part 1: Exploratory Data Analysis (20 points)
Explain the purpose of the analysis for your data, i.e., what are you trying to achieve?
Check for existence of NA’s (missing data)
Use appropriate plots for EDA, i.e., word counts for text data.
Do you need to scale your data or do you need dimension reduction? If so, perform a principle components analysis on a scaled, or not-scaled data depending on your needs.
Part 2: Clustering (20 points)
Develop a clustering algorithm for your data: Choose from topic models, k-means, k-medoids, hierarchical, or DBSCAN.
Explain your choices on model parameters, i.e. k, eps, minpts, and communicate your results.
Part 3: Mixture Models (20 points)
Apply a mixture model based clustering to your data.
Explain your choices on model parameters, and communicate your results.
Part 4: Deep Learning (20 points)
Apply a type of neural network algorithm to your data.
Explain your choices on model parameters, and communicate your results.
Part 5: Conclusion (20 points)
(10 points) Based on the purpose of your analysis stated in Part 1, which analysis did a good/better/satisfactory job? How do you think you can improve the analysis?
(10 points) What are your learning outcomes for this assignment? Please focus on your learning outcomes in terms of analysis, model interpretations, and R skills - it is up to you to include this part in your presentation or not.