EarthquakePrediction Using Bi-nomial Logistic Regression
Overview- The present invention relates to a method of using binomial regression，coupled with existing data and future sample data gathered by means of satellite fault monitors for the purpose of determining the variables that lead to the occurrence of earthquakes. The data is then paired and using regression analysis we identify the coefficients，using the least squares model，that are most useful in accurately predicting a future quake.
In a variant，the method comprises using empirical data，such as found on the USGS database，or collecting new samples by satellite，to determine a base model for why and when earthquakes occur. The method seeks to statistically evaluate all possible variables that lead to the sequence of events，which precede the actual slippage and reformation of the earth's crust. The author seeks to integrate electromagnetic，solar or other signals seeking to combine all possible variables that can be accurately measured and quantified. The author assumes that there may be multiple causes and variables，which when combined result in a seismic event. The author also assumes that inference can be accurately applied，by way of studying micro-quakes and using that resulting data to generate statistical models geared towards predicting future seismic activity on a larger scale.Current technology allows us to predict quakes within 10 seconds，the author teaches that either earthquakes lay outside other natural phenomenon or the data is incomplete or being misinterpreted. This model will run binomial regression analysis using Bayesian analysis to take new or unknown variables into account and input them into the existing model. We will seek to combine as many variables as possible so that a user can ultimately mitigate damage，in terms of the cost of human lives，that are now are judged highly at risk.If the current monitoring methods are deemed to be insufficient we address these in the scope of our work.
IP- US Provisional Application
Start Date- January 1st， 2018
Advantages to Current Method
· There is no accurate method currently in use
· Logistic Regression has been used but as a function of space/time and curve fitting shocks we seek to understand the earths crust as a holistic environment
· Current methods have limited the number of variables tested a mistake we will avoid
· Bayseian Theory has also been conceived of as a method to understand the interaction between two seismic variables，as this is not a thinking being we may utilize the Wald Statistic to some degree but our model would look more at bi-nomial logistic regression using the r coefficient to weight variables
· Use of satellites and lasers to track movements in the Earth’s crust as well as marine faults not currently being tracked because traditional tracking methods are inadequate
· All current research or work in this direction of predictive modeling has been confined to the laboratory and research papers we want to actively test our model in real-time
Business Model- Traditional Corporation and product development/marketing
Cost – Estimated as over $100 million USD
ROI- Would vary depending on effectiveness of method. Successfully extending the predictive ability by even 10 minutes would massively decrease the loss of human life saving billions of dollars. We estimate that this would command 20x cost or $2 Billion USD.