Using Big Data to Predict Marital Success and Divorce Rates

Not all marriages succeed / Photo by: Geolilli via Shutterstock


Big data may the solution to why almost half of all marriages ultimately fail as data scientists, relationship experts, and dating websites increasingly turn to big data to heighten the success and longevity of the marriage or relationships of their clients, according to Larry Alton, writing for Datafloq.

eHarmony’s Predictive Analytics

One company that does such a thing is dating website eHarmony, which integrated IBM Analytics into its matchmaking system. Since eHarmony started using IBM Analytics, it was able to make 3.5 million matches a day for its users and monitor the success rates of those relationships, according to data expert Kevin Small

Such matches have resulted in over 600,000 marriages or an average of 486 marriages per day and accounted for 5 percent of all new marriages in the US in 2014. 

The divorce rate for married couples is just 3.8 percent, since eHarmony rolled out predictive analytics into its matchmaking process, IBM Analytics reported. On the other hand, the divorce rate for couples using other dating websites is 7 and 8 percent for those who met offline.

To systematically match people, eHarmony uses 29 compatibility factors which were drawn up by the company’s founder, Dr. Neil Clark Warren. Among these factors are sense of humor, spirituality, sociability, and ambition.



eHarmony’s Compatibility Matching System

To match prospective couples, eHarmony employs a system made up of three parts.

Compatibility Matching. People who join eHarmony are asked about their preferences on such matters as distance, income, age range, religion, smoking, and drinking habits. Afterward, they are asked to fill in an extensive questionnaire consisting of 150 questions designed to extract their personality and psychological profiles.

One of the questions asked is this: If your best friends have to pick four words to describe you, which four words would be these? To answer the question, eHarmony applicants have to choose words contained in a list. Some of the words are modest, respectful, affectionate, caring and spontaneous.

The answers to the questions provide eHarmony with data on personality, values, attributes, and beliefs. To be able to give a compatibility score, the answers are run through complex mathematical equations and compared to an index. The index is the result of the marital satisfaction survey of 5,000 users. The scores of the most-highly satisfied couples are used in predicting new scores. If the scores of the new applicants are higher than the scores for the most-highly satisfied couples, they are considered compatible.

Affinity Matching. The compatibility matching system will also predict the possibility of communication between two people because even if they are compatible with each other it may be futile to match them for several reasons. These include the willingness of either of the two or both to communicate only with someone living within a certain distance or from a specific age category. The system will not match these people.

David Gevorkyan, eHarmony’s principal software engineer, said most of the time communication happens when the males are four to eight inches taller than the women. But the possibility of communication dives to zero when the woman is taller than the man. He added that the words used in the personality profile also influences the probability of communication. For the men, these words are perceptive, physically fit, passionate, intelligent, funny, and optimistic. For the women, these words are sweet, funny, ambitious, thoughtful, and passionate.

Food preference is also an important factor. For instance, vegetarians are more likely to talk with other vegetarians, and their communication rate is 44 percent above average.

The dating website also extracts information from images in matching people, such as the color of the hair and the eyes. Jonathan Morra, eHarmony’s director of data science, admitted that using images as a reference for attractiveness is very difficult and subjective. Morra confided that eHarmony had limited success with it but found more success with extracted features.

Each user has an average of 1,000 attributes and when combined together, the users have answered 4 billion questions.

Match Distribution. eHarmony uses machine learning to determine how many matches to send each day and the exact time of the day.


It's not too late to find compatibility with someone / Photo by beccalee via Pixabay


Predicting Divorce by Married Couple’s Tone of Voice

While the idea of using predictive analytics to look for a lifetime partner may not sit well with some people, researchers at the University of Southern California have taken predictive analytics to the next level. They have developed an algorithm that can predict whether a married couple’s relationship has improved or declined through the years. The algorithm makes use of the tone of voice that a married couple uses when speaking to each other.

Through the use of speech processing technology, the algorithm reduces married couples’ conversations into component parts -- acoustic characteristics such as volume, pitch, intensity, jitter, and shimmer -- and detects if a voice is shaking or warbling because of emotion. The algorithm’s success rate was 79 percent, which is more accurate than session notes provided by marriage therapists.  

Margarine Consumption as Factor in Divorce

That eating margarine can cause divorce is a misapplication of big dData, according to Forbes magazine. The issue started after the divorce rate in the state of Maine was wrongly correlated with the per capita consumption of margarine in the US.