Intelligent loan recommendation agent
An artificially intelligent loan recommendation agent and a method of making loan recommendations are disclosed. The artificially intelligent loan recommendation agent can classify applicants into traditional and non-traditional categories. Based on this categorization, the agent may use different information to make loan recommendations for the applicant. For non-traditional applicants, applicant rating information can be used, including ratings of an applicant’s professional skills obtained from consumer review sites. The agent can also conduct virtual interviews and analyze body language and speech to determine if an applicant may be lying.
1 . A method of detecting fraud in a loan application process using an artificially intelligent loan recommendation agent, the method comprising:
conducting a virtual interview with a loan applicant;
retrieving loan application information associated with the loan applicant;
receiving video information associated with the loan applicant, the video information being captured during the virtual interview using a camera;
receiving audio information associated with the loan applicant, the audio information being captured during the virtual interview using a microphone;
receiving biometric information and movement information associated with the loan applicant, the biometric information and movement information being captured during the virtual interview using a wearable device;
receiving environmental information associated with the loan applicant, the environmental information being captured during the virtual interview using sensors of a smart home;
receiving position information associated with the loan applicant, the position information being captured during the virtual interview using a smart phone;
using the video information to analyze body language of the loan applicant using machine learning algorithms to detect irregular patterns in the body language of the loan applicant that suggest fraud;
using the audio information to analyze speech of the loan applicant using machine learning algorithms to detect irregular patterns in the speech of the loan applicant that suggest fraud;
detecting fraud in the loan application process based on the analyzed body language and the analyzed speech, and based on determining when at least one of the biometric information, the movement information, the environmental information and the position information indicates fraud, the fraud being associated with a confidence score based on an amount of information used to detect the fraud;
wherein the biometric information and the environmental information indicate fraud when the wearable device detects increased perspiration of the loan applicant and the sensors of the smart home detect a cool temperature in the smart home;
wherein the movement information indicates fraud when the wearable device detects at least one of twitching and knee bouncing; and
pushing a recommendation to deny a loan to the loan applicant over a network to a loan provider when the fraud is detected with a confidence score that exceeds a threshold.
2 . The method according to claim 1 , wherein the environmental information further indicates fraud when the sensors of the smart home detect coaching of the loan applicant by another person in the smart home.
3 . The method according to claim 1 , wherein the position information indicates fraud when the smart phone detects a discrepancy between an actual location of the loan applicant and a claimed location of the loan applicant.
4 . The method according to claim 1 , wherein the using the video information to analyze the body language of the loan applicant comprises using a deep neural network to identify the irregular patterns in the video information.
5 . The method according to claim 1 , wherein the using the audio information to analyze the speech of the loan applicant comprises using a deep neural network to identify the irregular patterns in the audio information.
6 . A method of detecting fraud in a loan application process using an artificially intelligent loan recommendation agent, the method comprising:
conducting a virtual interview with a loan applicant;
retrieving loan application information associated with the loan applicant;
receiving video information associated with the loan applicant, the video information being captured during the virtual interview using a camera;
receiving audio information associated with the loan applicant, the audio information being captured during the virtual interview using a microphone;
receiving biometric information and movement information associated with the loan applicant, the biometric information and movement information being captured during the virtual interview using a wearable device;
receiving environmental information associated with the loan applicant, the environmental information being captured during the virtual interview using sensors of a smart home;
using the video information to analyze body language of the loan applicant using machine learning algorithms to detect irregular patterns in the body language of the loan applicant that suggest fraud;
using the audio information to analyze speech of the loan applicant using machine learning algorithms to detect irregular patterns in the speech of the loan applicant that suggest fraud;
detecting fraud in the loan application process based on the analyzed body language and the analyzed speech, and based on determining when at least one of the biometric information, the movement information, and the environmental information indicates fraud;
wherein the biometric information and the environmental information indicate fraud when the wearable device detects increased perspiration of the loan applicant and the sensors of the smart home detect a cool temperature in the smart home;
wherein the movement information indicates fraud when the wearable device detects at least one of twitching and knee bouncing; and
pushing a recommendation to deny a loan to the loan applicant over a network to a loan provider when the fraud is detected.
7 . The method according to claim 6 , wherein the detecting the fraud is associated with a confidence score based on an amount of information used to detect the fraud and the recommendation to deny the loan to the loan application is pushed over the network to the loan provider applicant when the fraud is detected with a confidence score that exceeds a threshold.
8 . The method according to claim 7 , wherein the confidence score is based on an amount of information used to detect the fraud.
9 . The method according to claim 6 , wherein the biometric information indicates fraud when the wearable device detects at least one of increased perspiration of the loan applicant and an increased oxygen level of the loan applicant.
10 . The method according to claim 6 , wherein the environmental information further indicates fraud when the sensors of the smart home detect coaching of the loan applicant by another person in the smart home.
11 . The method according to claim 7 , wherein the sensors of the smart home detect a lighting level of the smart home to thereby detect the fraud with a greater confidence score by considering the lighting level when detecting the fraud.
12 . The method according to claim 6 , further comprising receiving position information associated with the loan applicant, the position information being captured during the virtual interview using a smart phone, wherein the position information is used to detect the fraud.
13 . The method according to claim 12 , wherein the position information indicates fraud when the smart phone detects a discrepancy between an actual location of the loan applicant and a claimed location of the loan applicant.
14 . A method of detecting fraud in a loan application process using an artificially intelligent loan recommendation agent, the method comprising:
conducting a virtual interview with a loan applicant;
retrieving loan application information associated with the loan applicant;
receiving video information associated with the loan applicant, the video information being captured during the virtual interview using a camera;
receiving audio information associated with the loan applicant, the audio information being captured during the virtual interview using a microphone;
receiving biometric information and movement information associated with the loan applicant, the biometric information and movement information being captured during the virtual interview using a wearable device;
receiving position information associated with the loan applicant, the position information being captured during the virtual interview using a smart phone;
using the video information to analyze body language of the loan applicant to detect indicators that suggest fraud;
using the audio information to analyze speech of the loan applicant to detect indicators that suggest fraud;
detecting fraud in the loan application process based on the analyzed body language and the analyzed speech and based on determining when at least one of the biometric information, the movement information, the environmental information and the position information indicates fraud, the fraud being associated with a confidence score based on an amount of information used to detect the fraud;
wherein the biometric information and the environmental information indicate fraud when the wearable device detects increased perspiration of the loan applicant and the sensors of the smart home detect a cool temperature in the smart home;
wherein the movement information indicates fraud when the wearable device detects at least one of twitching and knee bouncing; and
denying a loan to the loan applicant when the fraud is detected with a confidence score that exceeds a threshold.
15 . The method according to claim 14 , wherein the environmental information further indicates fraud when the sensors of the smart home detect coaching of the loan applicant by another person in the smart home.
16 . The method according to claim 14 , wherein the position information indicates fraud when the smart phone detects a discrepancy between an actual location of the loan applicant and a claimed location of the loan applicant.
17 . The method according to claim 14 , wherein the detecting fraud uses machine learning algorithms including at least one of supervised learning algorithms, unsupervised learning algorithms, reinforcement learning algorithms, regression algorithms, neural network algorithms, support vector machines, decision trees, Q-learning algorithms, and clustering algorithms.
18 . The method according to claim 17 , wherein the machine learning algorithms detect at least one of the irregular patterns in the speech of the loan applicant and the irregular patterns in the body language of the loan applicant using at least one deep neural network.