Next, please find a description of the already confirmed modeling problems.
“Ranking of patients for dental home care”
The “Hospital Provincia Cordillera” (HPC) runs a pilot project for dental home care aimed at patients with severe dependency. HPC needs a model to determine preferences for patients derived from primary attendance regarding complexity, level of dependency, age, socio-economic situation, among others.
Company/Organization:CRS - Hospital Provincia Cordillera
“Identifying Influencer Campaigns in Twitter”
For this work, you are expected to determine relevant features and develop one or more models to detect influence campaigns on Twitter.
Influence operations have been around since the start of human societies but in recent years, due to the global nature of the Internet, influence operations now have a larger scale and scope than ever before. By exploiting social media, nation states, organized crime, and even individuals can shift public discourse to achieve an actor’s goal.
By using publicly released data from Twitter, research the features that are relevant to detect a coordinated influence campaigns on Twitter. With these features, develop a proof-of-concept model that can detect these influence campaigns. Statistical analysis should establish the importance of each feature and how the model can identify when an influence campaign is occurring.
“Coverage and staffing of offices for face-to-face attention”
The bank needs a model that for each one of its offices determines the optimal number of executives to meet the respective demand, differentiating between preferential and standard customers. Furthermore, a model should be developed to propose openings of new offices with the desired staff level.
“Recommendation Engine for Online Sales”
The Company needs a recommendation engine that maximizes sales in its virtual shop using the following information: the customers’ surfing behavior, their past purchases (online as well as in a store), socio-demographic profile, among others.
"Development of a Machine Learning methodology for image processing to detect and classify external retinal damage"
Abnormalities at the level of the external retina, driven by age-related macular degeneration (AMD) and hydroxychloroquine (HCQ) toxicity, may cause irreversible blindness. Currently, AMD is the leading cause of central vision loss in elderly people and HCQ toxicity has an estimated incidence of up to 7.5% among users of this medication. In both conditions, an early diagnosis is capital for the preservation of visual function.
Deep Learning has shown a good performance mainly with fundus images and optical coherence tomography (OCT) in the detection of ocular conditions such as: diabetic retinopathy, AMD, retinopathy of prematurity.
This project aims to develop and validate a Machine Learning methodology based on OCT images for the automated detection of external retina damage.
Preliminary results (1,812 images) have shown an appropriate identification of abnormal OCT scans through the determination of the shape, thickness, and reflectivity of the area of interest.
Company/organization: School of Medicine, Universidad de Chile
Modeling Week 2020