136 lines (135 loc) · 4.18 KB

SparseMatrixRecommender work plan

Introduction

This org-mode file has a work plan list and a unit test list that correspond (and hopefully reflect) the development of the Python package SparseMatrixRecommender.

Work plan [33%] [3/9]

Object structure

[X] Sparse matrices
- A dictionary of SSparseMatrix objects.
[X] Recommender matrix
[X] Tag type weights
[X] Pipeline value

Core functionalities [66%] [4/6]

Getters [100%] [4/4]

[X] Sparse matrices
[X] Recommender matrix
[X] Tag type weights
[X] Pipeline value

Setters [100%] [4/4]

[X] Sparse matrices
[X] Recommender matrix
[X] Tag type weights
[X] Pipeline value

Check inputs [100%] [2/2]

[X] Verify dictionary of tags
[X] Verify dictionary of items

Sub-matrix retrieval

[X] By accessing object’s dictionary of matrices
[C] By taking (main-)matrix columns

Recommend by profile [75%] [3/4]

The first non-trivial operation to make
[X] List profile
[X] Dictionary profile
[ ] SSparseMatrix profile
[X] Normalization of results

Recommend by history [75%] [3/4]

[X] List history
[X] Dictionary history
[ ] SSparseMatrix history
[X] Normalization of results

Creation functions [100%] [3/3]

Main pipeline creation functions [100%] [3/3]

[X] From matrices
[X] From data frame in long form
[X] From data frame in wide form

Cross tabulation [100%] [3/3]

Cross tabulation of data frames into SSparseMatrix objects.
[X] Two columns
[X] Three columns
[X] Dictionary of SSparseMatrix objects
- For specs that have
  - index
  - list of columns (tag types)
  - values

Categorization of numerical values

[X] Specified breaks
[X] Quantiles

LSI functions [75%] [3/4]

Global LSI functions [75%] [3/4]

[X] IDF
[X] GF-IDF
[X] Column sum
[ ] Entropy

Local LSI functions [100%] [4/4]

[X] Binary
[X] Log
[X] Frequency
[X] Row sum

Normalizer LSI functions [100%] [3/3]

[X] Cosine
[X] Row sum
[X] Abs max

LSI functions application

[X] Global weights calculation function (separate)
[X] Main sequence of steps

Representation [66%] [4/6]

[X] __str__
[X] __repr__
[X] To dictionary form
[X] From dictionary form
[ ] to_wl stub

Second wave functions [20%] [1/5]

[X] Classify by profile
- [X] Program
- [X] Test
[X] Filter matrix
[X] Retrieve by profile with search engine specs
- I.e. having the specs: should, must, must_not
[ ] Average row per segment
[ ] Top-K statistics facilitation

Third wave functions (extra) [100%] [1/1]

[X] Metadata recommender derivation / conversion
- [X] Using matrix multiplication
- [C] Using queries over the long form representation of the recommender matrix

Recommender algebra [0%] [0/5]

[X] Remove tag types
[ ] Row binding with a matrix
[X] Column binding with a matrix
[X] Join across with another recommender
- Concatenate horizontally
- (Column binding)
[ ] Join with another recommender
- Stack vertically
- (Row binding)

Composite design pattern application [%] [/]

Composite of recommenders

[ ] Composite class
[ ] Composite operation

Chain of responsibility recommender

[ ] Composite class
[ ] Composite operation

Combined SMR and LSA objects

[ ] Composite class
[ ] Composite operation

Unit tests [83%] [5/6]

The unit tests (so far) are made with small matrices in order to make easier visual verification (by humans.)

Basic functionalities [100%] [3/3]

[X] Creation
[X] Recommend by profile
[X] Recommend by history

Creation

Representation

Classification by profile

Metadata recommender creation

Recommender algebra

Documentation [%50] [1/3]

Usage messages

Usage message for each method.

Diagrams

[X] SMR workflows diagram (reused)
[X] SMR pipeline diagram for the Python class.

First version

Advanced use cases