Project Title: Time Series Forecasting Model and its JAVA Implementation
Primary Skill: Programming
Secondary Skill: Java
Additional Skills: Statistical algorithms for Time series model for forecasting the predictions based on large data are essential.
We need professional to develop the following Predictive model for modeling large data coming from Banking, finance, Insurance, Telecommunication and such other fields.
Project Description: We need professional to develop the following Predictive model for modeling large data coming from Banking, finance, Insurance, Telecommunication and such other fields.
Time Series Forecasting Model:
The set up for developing Time series forecasting model is : One response variable measured on a time-scale with several predictor variables (may be continuous or categorical). Usually the data size is large and it will be stored in EXCEL or ASCII format. Assume that column represents a response variable while Rows indicate the cases.
Develop a statistical algorithm for fitting and selecting ‘best’ ARIMA (p, d, q) (P,D,Q) _m [where m is the seasonal frequency] model to univariate time series consisting of one continuous response variable and several regressors (may be continuous or categorical). Assume that the input data is homoscedastic.
Write explicitly the computational steps involved in the fitting of the above model. Your algorithm steps should clearly indicate the tools used to handle large data. Provide details of iterative method that you are going to use to fit the model along with convergence criteria.
Develop a Java Code corresponding to the above algorithm.
Your program should not crash at extreme values or invalid input. Your code should have built-in checks to avoid crashes and or incorrect results. This has to be documented with suitable test cases.
Expected output: The output should produce best ARIMA model with estimated coefficients, p value, standard estimate of the coefficients and log likelihood according to either AIC, AICc or BIC value. The function should search over possible models within the specified Order (p, q, P, Q) constraints.
Using the final ‘best ‘model your program should generate forecast of response and confidence interval of the forecast value for specified period ahead.
Test case results to demonstrate how the program behaves at extreme values, Invalid inputs etc. Your program should work correctly for large data. Include at least one example to demonstrate working of the program.
Your program should select a proper subset of predictors. It should also rank them as per their percentage of importance in the model.
Model Validation should be done and the output should indicate the prediction accuracy and test of white noise residuals of the fitted model.
Time Line: Time to develop a statistical algorithm and its Java implementation should not exceed Two months.
Budget: $4000-$ 6000