Back to top
  • Introduction
  • Methodology
  • Project outcome & Visualisations
  • Summary

    • Hierarchical Clustering

    Hierarchical ClusteringDownload

    To map emotional distances across the forum posts, we applied a hierarchical clustering algorithm. This completely separated the conversations into two distinct groups. The first group contains clinical realities and physical burdens. Patients discuss chemotherapy, weight loss, and hair loss within this cluster. The second group gathers psychological struggles. It includes faith, sleep routines, and hobbies like motorcycles. Doctors usually target the clinical metrics within the first group. Patients dedicate massive mental energy to the second group to maintain normalcy.

    Intertopic Distance Map (UMAP)

    Intertopic Distance Map (UMAP)Download

    Looking at the Intertopic Distance Map, acute physical symptoms like mouth sores and nausea cluster tightly in the center. These form the universal core of cancer treatment. Lifestyle and coping topics disperse widely at the edges. Reading and faith represent highly individualised coping methods.

    Sentiment Polarity Temporal Evolution

    Sentiment Polarity Temporal Evolution line chartDownload

    Using a RoBERTa sentiment model, our team tracked emotional changes between 2005 and 2021. Among all categories, Topic 9 emerged as an anomaly. This topic covers healthcare, insurance, and politics. While clinical side effects maintain a stable negative baseline, Topic 9 shows extreme volatility. During 2015, the sentiment score for this category crashed to nearly -0.8. This plunge correlates strongly with external debates regarding healthcare coverage and pre-existing conditions. The sentiment rebounded sharply in 2018 and 2021 when policy uncertainties stabilised. Systemic and financial toxicity generates more severe emotional trauma than physical chemotherapy side effects.

    Emotional Spectrum bar chart

    Emotional Spectrum bar chartDownload

    The overall community sentiment actually heavily skews positive. Topics related to faith, hope, and community support drive this positivity. However, the data exposes specific hidden burdens. Topic 9 holds the lowest average score at -0.20. Deep distress stems from financial issues rather than the disease itself. Mucosal toxicities and gastrointestinal issues also return explicitly negative scores between -0.13 and -0.12. Chronic daily quality of life degradations exhaust patients.

    Codes

    https://drive.google.com/file/d/1fC6HUXtpjH7bK-6-dYz2IzbeSd6LCA_8/view?usp=sharing