{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"

Recommendation Systems

\nIn this modern world we are overloaded with data and this data provides us useful information. But it's not possible for user to extract information which interest them from these data. `In order to help user to find out information about product ,recommedation systems where made`\n\nRecommeder system creates a similarity between user and items and exploits similarity between user/item to make recommendations\n\n**`What recommeder system can solve?`**\n\n1. It can help the user to find the right product.\n2. It can increase user engagement\n * Say there's 40% more click on google news due to recommendation\n3. It helps item providers to deliver items to right user\n * In Amazon, 35% or more products get sold due to recommendation\n4. It helps to make contents more personalized\n * In Netflix most of rented movies are from recommendations","metadata":{}},{"cell_type":"markdown","source":"

Formulation of problem is often more essential than its solution, which may be merely a matter of mathematical or experimental skill

\n\nEDA, MODEL, Python, Xgboost, NN are fine but before that lets see `Why H&M even care about Product Recommendation??` \n\n`Example:`
\n* Giant Amazon makes more then `40%` of there revenue only by product recommendation according to maybe-2017 or so, if you want to attach monetary aspects to it, so we are talking about `Billions of $` here\n* in 2016 it was more then `42Billions $'s`\n\nNow leverage this Information and try to get some intution about `Why H&M even care about Product Recommendation??` \n\n\nInternally Product Recommendation can be done in many ways but talking about Amazon they do it in these ways:\n1. `Content Based Recommendation`\n 1. say you searched for Nikey_Vneck-T, ther are high chance you will like other T which looks sames as Nikey_Vneck-T (maybe different brand, color etc...)\n 2. Here one can use `Text Description` and `Image Description` i.e. `Text and Image Content` to recommend \n\n\n2. `Collaborative Filtering` based techniques
\nExample:
\nSay there are 3 users-($U_1,U_2,U_3$) and they checked items-$i$
\n* User-$U_1$ check item's ($i_1, i_2, i_3, i_4.....,i_m$)\n* User-$U_2$ check for item's ($i_1, i_3, i_7, i_6......,i_n$)
\nNow we have new user-$U_3$ cheking for item $i_1$, now if we recommend him item-$i_3$ ther are more chance user atlest will not dislike it, and we are doing this by observing item cheking patter of ($U_1, U_2$) \n\n\n\n`For any Business it becomes hard to share information which is Required for Collaborative Filtering` \n\n\nNow if you want to recap your recommendation system knowledge or want to understand more about it go with this \n## [link](https://www.kaggle.com/mukeshmanral/recommendation-system-basic-surprise-library)\nI will update this notebook in future, you can upvote it anyway","metadata":{}},{"cell_type":"markdown","source":"Now Problem Statement is clear:
\n`Based on a Quary Item, suggest other similar items to user` by Leveragig `Text and Image data`","metadata":{}},{"cell_type":"markdown","source":"`I can say, internally H&M is using both Content Based REcommmendation and Collaborative Filtering`
\n`What do you think please commen????`","metadata":{}},{"cell_type":"markdown","source":"What is in above link you might think, so this is a glim of it:\n\n

Types of Recommendations

\nThere are mainly 6 types of the recommendations systems :-\n\n1. `Popularity based systems` :- It works by recommeding items viewed and purchased by most people and are rated high.It is not a personalized recommendation.\n2. `Classification model based`:- It works by understanding the features of the user and applying the classification algorithm to decide whether the user is interested or not in the prodcut.\n3. `Content based recommedations`:- It is based on the information on the contents of the item rather than on the user opinions.The main idea is if the user likes an item then he or she will like the \"other\" similar item.\n4. `Collaberative Filtering`:- It is based on assumption that people like things similar to other things they like, and things that are liked by other people with similar taste. it is mainly of two types: a) User-User b) Item -Item\n5. `Hybrid Approaches`:- This system approach is to combine collaborative filtering, content-based filtering, and other approaches \n6. `Association rule mining` :- Association rules capture the relationships between items based on their patterns of co-occurrence across transactions.","metadata":{}}]}