In recent years, a huge amount of information is available on the internet and it is very difficult for the user to collect the relevant information. While purchasing any product also a lot of choices available and the user is confused about what to choose. This will be a time-consuming process as well. The search engine will solve this problem to some extent by but it will fail in giving a personalized recommendation. In this presentation, I will describe the different types and working of the recommender system how they gather the data, build recommender, generate recommendations from it, evaluate the performance and effectiveness of the recommender system. The further part of the presentation will describe how to build a movie recommender system using python.

1. Movie Recommender System
Using Artificial Intelligence
Oswal Shrutika Suresh

2. Introduction
Types of Recommended System
Implementation of Recommended System
Feature of benchmarking a recommended
Practical implementation of movie recommender
Conclusion
references

3.  In recent years recommended systems are all around us it becoming more and
more popular.
 Recommendation Systems filter and predict the rating, views, preferences that
any user has given to product, movie, books, news, avertisse, Song, social tags,
etc.
 It is one kind of information filtering system that produces a list of
Recommendation.
 By using the system’s algorithms, it will show accurate user preferences by
analysing a huge number of datasets.

5.  Content-based filtering is also called as cognitive filtering.
 It depends on the profile, preference added by the customer and the description
of products.
 The challenge is extracting all discrete details of every product available
Content-based filtering system need to face several issues like:
1) Some terms in the description of a particular product can be assigned
manually or automatically.
2) To choose an algorithm that will make the best recommendation in a
particular scenario.
3)The terms are chosen in such a way that we are able to compare the
item’s description and the user profile preferences in some meaningful manner.

6.  The Collaborative filtering is also called as social filtering.
 The key idea is the person will more likely to be agreed in the future if they had
agreed in the past in the evaluation of certain items.
 There is no need to strictly monitoring specific kind of information as required
in content-based filtering.
 It analyses similarities between the customer's interests and their behaviour and
finally serves recommendation list.
 Popular examples are Spotify, YouTube, Netflix, etc.

7.  There is a need for fast response and it should be scalable according to very
large datasets.
 To satisfy the primary approach of speed and scalability we develop a
model by extracting information from large datasets.
 Advantages :-
1) Scalability
2) Recommendation speed
 Disadvantages :-
1) Inflexibility
2) Quality of Recommendation

8. Clusters of users and projects
are built upon the basis of the
user’s rating, user's interest,
project attribute vectors.
K Nearest Neighbour (KNN)
algorithm is used to implement
this clustering model.
• Some small number of hidden
factors are taken into
consideration for determining the
attributes or preferences of the
customer.

9. This is an extension to matrix factorization.
It makes use of Multi-layered neural nets including embedding layers

10.  This technique utilizes entire datasets to generate recommendations.
 The customers who purchase similar kind of items or the customers who give a
rating to different items similarly are knowns as a neighbour
 The systems find this kind of neighbours by applying some statistical technique
 Types of Memory-based Filtering:-
a) User-item Filtering
b) Item-item Filtering

11. Advantages :-
 Need to develop a particular model.
 We are using the entire database at every new prediction it is very easy to update the datasets.
 Quality of recommendation is good.
 Simple algorithm is used so easy to implement in any situation.
 Disadvantages :-
 Very slow process of prediction as it requires the entire database to be in memory every time.
 Memory requirement is more.
 It does not generalize the dataset at all.

12.  The solution to the Content-based filtering problem and Collaborative filtering problem is
collaboration via Content which is also called a Hybrid approach.
 User profile is constructed not only by the rated items but also by the content of item.
 There is a weight assigned to each term which indicates the importance of that term.
 It is able to give recommendations outside of normal user environment based on the
experiences and impressions of another customer.

14.  The first task is to collect a large amount to data for processing.
 This gathering of relevant data process involves multiple users and item information in the form of user
behaviour, user interest, item rating, discrete item attributes.
 It starts with data filtration and structuring.
 User data with ratings and item attributes with keywords are given as an input.
 It also takes design recommendation interface model outcomes as an input parameter and finally, the
updates will be passed to the recommendation model.

15.  generate a list of recommendations and send it to the user via a user interface, any other social networking
sites or through advertising.
 Data scientist always try to recollect this data for the performance evaluation of recommender systems.
 To measure whether your recommender is up to the mark or not you need to do a survey

16.  user preference
 Prediction Accuracy
 Coverage
 Trust of user
 Confidence
 Novelty
 Diversity
 Risk Factor
 Robustness
 Utility
 Privacy

17. import numpy as np
import pandas as pd
ratings_data = pd.read_csv("D:Datasetsml-latest-
smallratings.csv")
ratings_data.head()
movie_names = pd.read_csv("D:Datasetsml-latest-
smallmovie_.csv")
movie_names.head()
movie_data = pd.merge(ratings_data, movie_names, on='movieId')
movie_data.head()
movie_data.groupby('title')['rating'].count().sort_values(ascending=F
alse).head()
ratings_mean_count =
pd.DataFrame(movie_data.groupby('title')['rating'].mean())
ratings_mean_count['rating_counts'] =
pd.DataFrame(movie_data.groupby('title')['rating'].count())
ratings_mean_count.head()

18. import matplotlib.pyplot as plt
import seaborn as sns
sns.set_style('dark')
%matplotlib inline
plt.figure(figsize=(8,6))
plt.rcParams['patch.force_edgecol
or'] = True
ratings_mean_count['rating_coun
ts'].hist(bins=50)

19. plt.figure(figsize=(8,6))
plt.rcParams['patch.force_edgecol
or'] = True
ratings_mean_count['rating'].hist(
bins=50)

20. user_movie_rating = movie_data.pivot_table(index='userId', columns='title', values='rating')
user_movie_rating.head()
forrest_gump_ratings = user_movie_rating['Forrest Gump (1994)'] forrest_gump_ratings.head()
movies_like_forest_gump =
user_movie_rating.corrwith(forrest_gump_ratings)corr_forrest_gump=pd.DataFrame(movies_like_forest_gump,colu
mns=['Correlation'])corr_forrest_gump.dropna(inplace=True) corr_forrest_gump.head()
corr_forrest_gump.sort_values('Correlation', ascending=False).head(10)
corr_forrest_gump = corr_forrest_gump.join(ratings_mean_count['rating_counts']) corr_forrest_gump.head()
corr_forrest_gump[corr_forrest_gump ['rating_counts']>50].sort_values('Correlation', ascending=False).head()

22.  In this paper we have seen various benefits of the recommended system.
 Also we have studied different types of recommenders with their advantages and
disadvantages.
 By using this kind of recommender system it is easy to provide the suggestions to
the customer so that they can choose a product according to their area of interest,
preferences.
 we have gone through various phases of implementation starting from gathering
the huge data set to generating real time recommendation to particular customer.

23.  Sarika Jain , Anjali Grover , Praveen Singh Thakur , Sourabh Kumar Choudhary; “Trends,
problems and solutions of recommender system,”
 [2] Bogdan Walek , Petra Spackova; “Content-Based Recommender System for Online Stores
Using Expert System”. 2018 IEEE First International Conference on Artificial Intelligence
and Knowledge Engineering (AIKE)
 [3] Ruchika, Ajay Vikram Singh, Mayank Sharma; “Building an effective recommender
system using machine learning based framework”; 2017 International Conference on Infocom
Technologies and Unmanned Systems (Trends and Future Directions) (ICTUS)
 [4] Kunal Shah, Akshaykumar Salunke , Saurabh Dongare, Kisandas Antala; “Recommender
systems: An overview of different approaches to recommendations”; 2017 International
Conference on Innovations in Information, Embedded and Communication Systems
(ICIIECS)