Embedded Files
Business Problem:
Business Problem:
You are working in the Human Resources (HR) department for Redwood Financial, a Chicago financial services company focused on individuals' investments and retirement planning. Your task is to do a data-driven analysis of the top 500 cities in the country, coming up with a recommendation of the top five cities for opening new Redwood Financial offices. Your work must be captured in a custom GPT to help persuade existing employees to move there. To do this, you will extract, merge, transform, and analyze five location-related datasets (on the course website) to develop a scoring formula to rank-order the top 500 cities in the country.
The last time your company did something similar, it spent a significant amount of money creating paper booklets to persuade employees to consider a move to one of the new target cities. This time, instead of the cost and effort of creating booklets, Redwood Financial employees will interact directly with your custom GPT as they explore whether they are interested in moving to one of the five new cities. Your created custom GPT should make a compelling case for why each location is desirable, considering factors such as standard of living, housing costs, recreational opportunities, schooling, culture, safety, weather, the character of each place, and health factors. Your custom GPT should also explain the process used in arriving at its recommendations.
You will present your custom GPT to Redwood Financial partners (well, to CS 218 students and staff that you should treat as if they were Redwood Financial partners), describing your decision-making process and the steps involved in creating your custom GPT, giving a short live demo of it in action.
Steps:
Steps:
Bring to this project the best of your own thinking, along with the best help you can get out of Google Colab/Gemini and ChatGPT. You will complete the individual portion of this project by completing the following steps as homework in the next several classes.
- Select Employee Types and Features
Brainstorm and come up with at least four different types of Redwood Financial employees who will be the target audiences for your work. For each type of person, make a list of location characteristics they are likely to care about. For instance, one group might care more about school options, another about recreational options, and yet another about environmental safety.
Then go through the five location-related datasets (on the course website) and identify at least 15 different features that will likely be helpful, choosing at least one from each dataset. For each feature, describe how it might be relevant to one or more of the four target audiences you previously identified.
Create a google doc with a description of the different types of employees, the different features you selected from the five datasets, and how each feature does/doesn't apply to each of the employee types. Make sure you set permissions to allow us to see the google doc!
Submit your google doc link using the form on the course schedule.
2. Practice Extracting and Combining Dataset Features, Displaying Heatmap
2. Practice Extracting and Combining Dataset Features, Displaying Heatmap
Use Google Colab and Gemini to practice extracting fields from multiple existing datasets, merging them into a single new dataset. You may find it easier to do this in multiple steps.
First go to the course website Datasets page and select 4. Geographic Location Data. On the resulting page, open up the README file, and follow the directions to convert counties.json into counties.csv
Then find, extract and combine the following features, for each of the 500 largest cities listed in 500_Cities_2019_CDC:
Percent of population with teeth lost (from 500_Cities_2019_CDC)
Water area (from Income_by_City)
Cost of living (from Cost_of_living_index_advisorsmith)
Employment (from counties.csv that you create from counties.json)
Once you have created a new dataset containing the above four pieces of information for each of the top 500 cities, create a heatmap showing the correlations between them. Then describe what this heatmap reveals, in your own words.
Create a google doc with a link to your created dataset. Make sure permissions are set to allow us to see both the google doc as well as the dataset! Include a description of how each of the above four features would or would not be useful in identifying the top cities.
Submit your google doc link using the form on the course schedule.
3. Extract and Combine Features, Creating and Applying Rank-Order Formula
3. Extract and Combine Features, Creating and Applying Rank-Order Formula
Refine your final set of at least 15 features you previously chose from the five location-related datasets, with at least one feature chosen from each dataset. Extract and combine these features to create a new dataset containing only these features.
Then create a feature-weighting formula in which each feature contributes a percentage to rank-order the 500 cities. For example, if each feature contributed exactly the same amount (probably a poor choice), then each of the 15 features would be worth 6.67%. The sum of all the feature percentages should be 100%.
Apply your weighting formula to each city's feature values to obtain a score. Sort the dataset by this score, revealing the top five cities.
Create a google doc with your list of 15 features and your selected weights for each one. Explain your process and reasoning for choosing the weights as you did. Include a link to a google sheet with your rank-ordering of the 500 cities, with the best cities at the top. Be sure permissions are set so we can access it! In your google doc also list the top five cities.
Submit your google doc link using the form on the course schedule.
4. Create Custom GPT
4. Create Custom GPT
Create a custom GPT containing:
An uploaded text file describing the process you’ve used so far in what you have created for Project 2.
Your created dataset
Additional information about your top five cities into your custom GPT, to make the user experience as compelling as possible.
Capture your work in a Google Doc, including: 1. The link to your custom GPT and 2. A sample of what it looks like when running.
Submit your google doc link using the form on the course schedule.
Project Individual Submission:
Project Individual Submission:
On your website, create a page for project 2, if you have not already done so.
On that page capture the work you have done so far for this project, from the above steps. Do not simply have links to the above work done. Either copy/paste all the contents onto your new project 2 page, or embed the individual google docs into that page, with descriptive text for each section. This content must be your individual work, with the possible exception of the Custom GPT that your group may have collaborated in creating.
Submit the link to this published web page on your website using the form on the course schedule.
Grading
Grading
The project deliverables will be graded out of 5 * 3 = 15 points. Consider each of the preparatory pieces of work that should be present on your website:
Selected employee types and features
Results from practicing extracting and combining dataset features, displaying heatmap
Description of final set of extracted and combine features, with weighting formula explained and applied, revealing top cities.
Created Custom GPT, with description and sample of it running
Website STAR story as an organizing framework for the above work
Each of the above components of work will be assessed using the following evaluation rubric:
0 points: Not done
1 point: Significantly incomplete, or just a link and the document itself is not presented
2 points: Almost everything is there.
3 points: Complete and full descriptions. Includes appropriate visual elements.
For the custom GPT, it is complete (includes sample output) and persuasive.
For the website, it has all components present on the page, with descriptions,
structured as a STAR story.
Presentation Grading:
Presentation Grading:
The presentation itself is worth 3 * 5 = 15 points and will be assessed on the following:
All group members participate equally
Presentation is professional in preparation, appearance, and delivery
STAR story components are clearly presented
Feature selection, Ranking formula weights, and final top city selections are explained both intuitively as well as based on data
Custom GPT demonstrates the process and reasoning used in the work, and is persuasive
Each of the above components of work will be assessed using the following evaluation rubric:
0 points: Not done
1 point: Weakly present
2 points: Present
3 points: Strongly Present
4 points: Extra credit: Exceptional, far above and beyond expectations
Peer Feedback
Peer Feedback
Peer feedback from those in the room will count for an additional 5 points (out of 35 points total).
(This was added 3/3, FYI)
Overall Points:
Overall Points:
The total Project Individual Submission and Presentation scores will be summed and then scaled down to the total possible points for the project, recorded in the course Grades spreadsheet.
Page updated
Report abuse