quiz 1

시나리오 1:
당신은 고객에 대한 정보를 수집하여 표 형식의 데이터세트에 저장하는 프로젝트를 진행하고 있습니다. 프로젝트의 목표는 고객의 갱신 여부를 예측하는 모델을 만드는 것이며, 모델은 계정 담당자가 사용할 수 있도록 해석 가능성이 높아야 합니다.

고객의 갱신 여부를 예측하는 데 적합한 모델 유형은 무엇입니까?

Logistic Regression, Neural Network, Decision Trees
Since this is a binary classification problem, many machine learning techniques will work for the problem. However, of these three model families, there are really only two that meet our needs and one that is likely ideal. Let's continue to narrow down our decision. 

2. 위에서 선택한 모델 유형 중에서 시도할 모델 유형의 범위를 좁히는 데 도움이 되는 주요 제약 조건은 무엇입니까?
The aility to interpret the model (모델 해석 능력)
Model interpretability is an important constraint on our model choice, and unfortunately makes neural networks less-than ideal for our problem. That leaves logistic regression and decision trees as interpretable models we can use for binary classification problems in this scenario.

3. For the two remaining classes of models, what aspect of the problem suggests which model family we should try first?

나머지 두 가지 모델 클래스에 대해 문제의 어떤 측면이 어떤 모델 계열을 먼저 시도해야 하는지 제안합니까?

-> The tabular structure of the data
Decision trees do an excellent job with tabular data, are relatively interpretable, and are well-suited for binary classification problems. Logistic regression has many of these properties, but like neural networks, does not perform as well in general on tabular data.

Scenario 2:

Your organization has asked for a model that prioritizes customer tickets based on how angry the customer appears to be. The model will take customer emails as input and will output a score from -1 for very upset to +1 for very happy to describe how upset or happy the customer is. 귀하의 조직에서는 고객이 얼마나 화를 내는지에 따라 고객 티켓의 우선순위를 정하는 모델을 요청했습니다. 모델은 고객 이메일을 입력으로 받아 매우 화가 난 경우 -1점, 고객이 얼마나 화가 났거나 행복한지 설명하면 +1점으로 점수를 출력합니다.

1. What aspect of this problem makes it well-suited to using neural networks?
Neural networks do well with natural language processing tasks

-> This is a task known as sentiment analysis and the ability of neural networks to handle natural language processing tasks makes them ideal for the problem.