In this chapter, restricted Boltzmann machine-driven (RBM) algorithm is presented with an enhanced interactive estimation of distribution (IED) method for websites. Indian matrimonial websites are famous intermediates for finding marriage-partners. Matchmaking is one of the most pursued objectives in matrimonial websites. The complex evaluations and full of zip user preferences are the challenges. An interactive evolutionary algorithm with powerful evolutionary strategies is a good choice for matchmaking. Initially, an IED is generated as a probability model for the estimation of a user preference and then two RBM models, one for interested and the other for not-interested, is generated to endow with a set of appropriate matches simultaneously. In the proposed matchmaking method, the RBM model is combined with social group knowledge. Some benchmarks from the matrimonial internet site are pragmatic to empirically reveal the pre-eminence of the anticipated method.
Top1 Introduction
In the late 1990s, Indian matrimonial websites become increasingly important in the marriage market and have millions of participants (Fritzi-Marie, 2013). In the early days, uncertain moves are made through Artificial Intelligence (AI) in the matchmaking business and as machine thinking expands traction. The AI and machine learning models are together applied in the recent Matrimonial sites (Sharma, 2018). At present most of the matchmaking sites entitle to practice intellectual matchmaking algorithms that help to pair and suggest the user profiles constructed on a variety of parameters: ones perceptibly filled in by the person and ones collected through the machine learning model. In our daily lives, machine learning algorithms influences all manner: from cooking recipes to the best route to reach to work, which movie to watch or music to play next, what news up on social media, what next to buy in online shopping. Surprisingly, machine learning algorithms also determine who can be the best life partner. Of course, the algorithm is based on the furnished information and may go wrong if the information is false. Use deep learning techniques to measure the appeal of an individual.
India is one of the highest population countries that all we know at the same time it is the only country which has the largest young generation people with less than 25 years old. So, the data-driven procedures are applied to India’s matrimonial giants. Later in 2015, The Matrimony.com group has introduced a Matchmaking Algorithm called MIMA (Matrimony.com’s Intelligent Matchmaking Algorithm) (Sharma, 2018), which give immediate recommendations on profiles suitable to the user by applying techniques of machine learning and arithmetical rules (Jha, 2015). Subsequently launching MIMA matrimonial websites provide substantial growth in end-user-to-end-user metrics and these metrics have a balanced boost with incessant enhancements to the algorithm. As the result, the algorithm doubles the business and the users are attracted by MIMA recommended profile that significantly increases to “two-digit user finding matches” and it leads to marriage. Finally, the MIMA algorithms improve the business with great customer satisfaction.
Sometimes the end-user leaves certain information unfilled that time ML techniques are used to include those inferential data on a profile. That possibly related to the income of the person, for example, whether the person is well travelled or not. The thing to notice here is income level is important criteria in matchmaking but it cannot be asked directly and set as a mandatory input detail. Artificial Intelligence and machine learning can stumble at times. Matrimonial fraud is a serious problem; a fraudster may use matrimonial sites to ploy and cheat several individuals. The ‘Trust Badge’ system, verified social media profile or government ID, along with the method of monitor the person’s social network activities may work best (Sharma, 2018).
India’s first matrimonial website powered with AI is Betterhalf.ai (Rai, 2019). On this website, matchmaking is done through a personality test at the time of user registration using various types of questions and options. A LinkedIn account is mandatory to register and users’ official email also verified along with their profile to eliminate fraudulence. A set of 16 questions helps to scale the user preferences and provide analysis of 6 dimensions of a user’s personality, such as intellectual, emotional, social relationship, physical and economic values. Update the forecasting scores of personalities depending on the privately fetched opinion about a person acts from the social network, which further refined using gamified questions on personality. A rating scale called a Likert Scale which uses a point scale as a replacement for imprecise questions or simple yes/no responses. In general, 5 or 7 measuring points used in this scale to identify exactly what a customer or visitor feels about a profile or website.