Merge pull request #191 from The-Strategy-Unit/184_yiwen_AI_presentation

incorporates Sheilas suggestions

presentations/2024-10-10_what-is-ai-yiwen/index.qmd

@@ -56,12 +56,12 @@ Data is consistent within each column in the table.
 
 The length of each sentence is different - data not consistent.
 
-## A simple approach to classifying data
+## A simple approach to classifying data: KNN
 
 :::: {.columns}
 ::: {.column width="50%"}
 
-Clustering algorithms like K Nearest Neighbours are on the more basic end of the scale, requiring very little computational power.
+Clustering algorithms like K Nearest Neighbours (KNN) are on the more basic end of the scale, requiring very little computational power.
 
 :::
 
@@ -72,25 +72,38 @@ Clustering algorithms like K Nearest Neighbours are on the more basic end of the
 :::
 ::::
 
-## More complex approaches to classifying data {.small}
+
+## A simple approach to classifying data: Decision Tree
+
+![Example of a decision tree](https://www.researchgate.net/profile/Dan-Andrei-Sitar-Taut/publication/26635430/figure/fig1/AS:883068449607681@1587551333416/Decision-tree-classifying-patients-with-coronary-artery-disease.ppm)^[https://www.researchgate.net/publication/26635430_Using_Machine_Learning_Algorithms_in_Cardiovascular_Disease_Risk_Evaluation]
+
+## There are many different models out there! 🥴 {.small}
 
 ![Complex diagram showing a decision tree for choosing the right estimator for different machine learning problems](https://scikit-learn.org/1.3/_static/ml_map.png)^[https://scikit-learn.org/1.3/tutorial/machine_learning_map/]
 
-::: {.notes}
+## What makes a model simple or complex?
 
 - There are dozens of different algorithms out there
-- Complex neural networks can take days to train if the dataset is large enough
 - Each algorithm has different strengths and weaknesses
-
-:::
+- What makes a model simple or complex is the amount of computational power required and how much the model needs to "learn" - how many parameters there are
 
 ## Is the input or the computation complex?
 
-The University of Leeds has helped **train an AI system** called Optimise, that looked at **health records** of more than **two million people**.
+"We used UK primary care EHR data from 2,081,139 individuals aged ≥ 30 years... 
 
-...
+We trained a random forest classifier using age, sex, ethnicity and comorbidities (OPTIMISE)."^[Nadarajah, Ramesh, et al. "Machine learning to identify community-dwelling individuals at higher risk of incident cardio-renal-metabolic diseases and death." Future Healthcare Journal 11 (2024): 100109. https://www.sciencedirect.com/science/article/pii/S2514664524002212]
 
-Of those two million records that were scanned, more than **400,000 people** were identified as being high risk for the likes of heart failure, stroke and diabetes.
+## How do we know if a model is good?
+
+Beware of metrics like "accuracy"
+
+If only 1% of the population is at risk of heart disease, and the model is a broken model that never ever predicts someone is at risk, this model would still have 99% accuracy, because it would be right 99/100 times!
+
+Carefully consider what is important to measure, given the context.
+
+::: {.fragment}
+In the scenario above, a measure like "recall" would be more useful.^[https://www.youtube.com/watch?v=qWfzIYCvBqo]
+:::
 
 ## Pros and cons of simple "A.I." approaches
 
@@ -117,6 +130,7 @@ Cons:
 ## 🚩 Issues to look out for 🚩
 
 - How complex is the input, or the computational approach?
+- How is the model's performance measured?
 - Does the model get updated?
 - Where did the data come from?
 - Have issues of bias or ethics been considered?