REPORT # Business Questions after attending [[Course attended - Data Analytics Begins with me NUS]] I attended a course, "Data Analytics Begins with Me - NUS," on 18 Feb 2022 to understand how the data analytic process works. We learned about the data analysis process and the different machine learning methods in the Course. We are introduced to an easy to use, drag-and-drop, free software, "Orange", that we can easily use to analyse our data. In our work, we generate a massive amount of data daily. It will be a waste if we do not analyse it and turn them into information. Information turns into knowledge, and that is powerful to guide businesses. ## The different types of analytical methods. Trainer shared that there are four kinds of analytical methods. Descriptive, Diagnostic, Predictive and prescriptive. (Slide below). I think scientific research uses descriptive-analytical methods. It describes what is happening in a particular phenomenon. ![[Pasted image 20220221154726.png]] ## Data Analytics Process The following diagram illustrates the data analytical process. Very easy to understand. We first begin with a business question (i.e. What are we curious about? what do we want to solve? e.g. How to increase sales by 20%, or what if "What factors are most important in predicting relapse?") ![[Pasted image 20220221155306.png]] Some business questions I am curious about are as follows: ## Understanding our Customers Borrowing idea from customer Segmentation, where the commerce platform (like Shoppee) wants to under its consumer segmentation. **I wonder, what are the different segment classifications of patients (i.e. education, DUP, level of compliance, frequency of medication changes? ...)** (Cluster Analysis) 1. Understand who our patients. (Profile of customers) are 2. Then, can we target different patients with specific interventions. Maybe they have different needs? ![[Pasted image 20220221154133.png]] ![[Pasted image 20220221153824.png]] ![[Pasted image 20220221153402.png]] **Curious: What profile of patients are diagnosed with specific disorders?** - Is there a pattern? ## Psychopathology? **What predict or correlate to "x", It could be treatment outcomes, risk level, compliance level** (Random Forest. Tree-based Models) 1. Can we build a model to calculate risk levels more objectively? (if it's a particular age group, sex, specific psychosocial circumstance). What are the factors that predict an outcome? compliance level? Risk levels? 2. Number of caregivers in the family? 3. Patient to clinicians - correlation to frequency of admissions, relapses? What factors have higher weightage? (The machine automatically calculate it) ![[Pasted image 20220221153420.png]] ## Service / Manpower Resources **What is the distribution of patients in the different geographical regions, where do we have more patients?** (Simply superimposing postal codes onto a geographical map. And descriptive analysis) 1. Then, we can relocate the number of CMs to those regions *Performance related* **Among all the workers, who have higher relapses, higher risks, and harder to TOC. And find out what kind of patients profiles they hold. Any pattern we can see from there?** Can we provide targeted supervision, training, and equipping for the worker? **Reflection** 1. I find that Data Analytics is not scary. It simply applies statistical sense, machine learning (and we have tools for non-coders), and our domain expertise to make sense of data. 2. Relating to HOPE-S research, digital sensors will generate a high volume of new data (HRV, Sleep, activities, ambient lightings), much of which we do not know how to make sense of at this point. "We don't know what we don't know." So I am curious what our data got to tell us. **Recommendations** 1. Set a business question 2. Prepare our existing data (Merging and tidying all the data we have, in different places) 3. Go through data analysis. !y[[Report data analytics course.pdf]]