lathomas42 / recomando

Basic Recommendation engine that incorporates different techniques in Machine Learning and statistics.

#Reccomendation Engine Test

Logan Thomas

Purpose:

• To build a basic reccomendation engine using python and c++
• database will be stored using sql

##Intended Databases ###Purchase

• Each entry is an purchase, this db will hold all purchases made. This information can be used for both personalized and unpersonalized engines.
• Each entry will have the form UserID, ItemID, PurchaseID
  • UserID: Each User's unique Identifier.
  • ItemID: Each Item's Unique Identifier, note that for a purchase in which a user buys 3 items, 3 entries get added to the Purchase Database.
  • PurachseID: Each time a user checksout that is given a unique purchase Identifier, as explained before, if a user buys 3 items at once, 3 entries are added however they all get the same PurchaseID.

###Ratings

• Each Entry in this database is a rating. Ratings can be used to assist personalized Recommendation engines.
• Each entry will have the form UserID, ItemID, Rating
  • Rating: A number between 1-10 telling how much the user liked the item

###Items

• Each entry is an item, this keeps track of some characteristics of the item
• Each entry is formatted: ItemID, Name, Price, CategoryID
  • Name: Name of the item
  • Price: Self explanatory, price of the item
  • CategoryID: Unique identifier of the category the item belongs in

###Users

• Each entry is a User. Information contained in this assists in personal engines.
• Each entry is formatted: UserID, ZipCode, PersonalPrefs
  • ZipCode : If user puts in location then this will be their zip code, distance measures can be made from this information
  • PersonalPrefs : this can be more than one field. If user is prompted to give information about themself this is where that is stored. This can be used for personal engines.

##Intended Classes

###DataBaseManager

• DataBaseManager provides a python interface for dealing with the sql(probably csv as the start, sql later) database
• will handle adding, removing, and querying entries from all the databases.

###Recommender

• Recommender is a base class meant to be inherited and overwritten.
• Will provide basic functions and attributes common to all Recommenders.
• Options to recommend by item, user, or both.
• Will work with DataBaseManager to provide meaningful recommendations.