An Item Response model based problem select - test - result project

For the core theory and the basic of IRT, pls refer to this site

Parts

Planning 3 parts: item parameter estimation (Python), API server (Node.js), and frontend (AngularJS)

• Item parameter estimation calculated all the needed parameters (chapterId, topicId, difficulty, discrimination, and guessing index), this part including 3 steps:

• Step 1: Given an ObjectId of a chapter, pick a specific number (cur. 50) of problems in its topics, and return the userIDs who did all the problems in a topic (Done)
  To do in next sprint: optimizing processing time using multiple processes
• Step 2: Estimate the difficulty, discrimination, and guessing index per problem (i.e item) (Done)
  • 2.1 Given a list of problems and its relevant testers, return the probability of right answer for each group of testers (done)
  • 2.2 Given the probability of right answer for each group of testers, return problem distinction and difficulty (done)
• Step 3: Store the data into a new db collections (Done)

• API Server

• Step 1: Design website instantiation and data structure
• Step 2: Examinee estimation algorithm (incl. examinee ability estimation, item response and re-response algorithm, Python or Node?)
• Step 3: Build APIs according to ds and algorithm

• Frontend

• Step 1: Basic framework and working progress
• Step 2: Data manipulation
• Step 3: Beautify