A predictive model for the Kidnapper

It was 11.00 PM night and I set the alarm clock for 4.45 AM morning, tossed my phone on the nightstand and slept by thinking that the alarm clock will wake me at dawn with the reminder for the morning run. As evident, the convergence rate difference of the algorithms becomes more apparent over time and similar to that my sleep was turning into more depth over time.

Time was travelling very fast to the past so that I can rub my eyes and can peer at the blurred figures of the clock in the morning. Meanwhile, I started dreaming and a sequence of a flash of a dream passed through my mind. And, in my dream, I knew that kidnappers kidnapped me and my friend and asked for the predictive model solution related to kidnapping rather than a ransom.

Problem statement of Kidnappers

Kidnappers are very optimistic about the amount worth $500 Million per year from the kidnap and ransom industry estimated worldwide in 2010. They explained to us the problem statement related to kidnapping and wanted to move from Business Intelligence to Predictive Analytics.

The traditional approach of kidnappers:

1. When did person “A” last visited the kidnapping spot?
2. How many persons were kidnapped in a city over the past month or year?
3. How much revenue did kidnapping gang “X” generate last year?
4. What is the ransom amount trend for a particular incomer or designated person over the past 12 months?
5. From which cities and countries did they get the highest ransom?
6. Which city and countries were safe for this activity?
7. Who are the most profitable group over the past 12 months?

But now they want to predict an outcome with a given input data by validating and evaluating a model so that they can escape from failure, penalty and criminal offence. They explained to us several use cases with an optimistic attitude and most of the uses cases were suited either on Regression Modelling or on Classification modelling.

Discussed use cases are mentioned below:

Classification Modelling:

They had a huge amount of historical or past data related to kidnap and ransom and wanted to predict the label (class) of the unlabeled examples.

1. Is kidnap to a particular city is safe or unsafe from government officials and legal actions?
2. Is the site suitable for the victim to hide?
3. Will a particular type of person call the police or pay the ransom?
4. Classify and label the type of kidnapping like child abduction, Shanghaiing, Fetal abduction, Extraordinary rendition, hostage, etc.

Regression Modelling:

They wanted to forecast and find the causal effect relationship between the variables. Their main purpose was to create a predictive score.

1. Relationship between kidnapping and number of punishments by law.
2. Ransom amount prediction with different targeted cities.
3. Forecast ransom amount for Quarter 4.
4. Chances to get caught by police based on location, type of officer and season.
5. Which city will be the most valuable and successful for the ransom.

While explaining the use cases kidnappers were now sitting drowsily and snoring. A quiet investigation revealed all doors were open and the sound of snoring from the kidnappers enforced us to run. Finally, we escaped from the house of kidnappers named “Ariel Castro”. The alarm clock buzzed insistently and I reached over to slap the snooze button, squinting at the iridescent hands and it was 4.45 AM morning.

Now the predictive analysis is everywhere, in the mind of each and every person, every business whether that is good or bad. It is an accurate demand to optimize your resources as you look to make decisions and take actions for the future.