1. try to find df(t), the integral get y(T) = y(t) + ing_t^T df(u)
2. never use for prediction, as we derive the pricing under risk neutral measure, it won't transfer to physical world.
3. meaning are a) pricing new products, b) risk management by knowing the sensitivity, c) secure quants' jobs.