Sequential Monte Carlo Filters with Parameters Learning for Commodity Pricing Models

In this article, an estimation methodology based on the sequential Monte Carlo algorithm is proposed, thatjointly estimate states and parameters, relationship between prices of futures contracts spot primary products determined, evolution volatility historical data market (Gold Soybean) are analyzed. Two stochastic models for parameters considered, describe physical measure (associated with price) risk-neutral markets to futures), price dynamics in short-term through reversion mean while that long term futures. Other characteristics such as seasonal patterns, spikes, dependent volatilities, non-seasonality can also be observed. methodology, a parameter learning used, specifically, three algorithms (SMC) state space modelswith unknown parameters: first method considered particle filter sampling importance resampling (SISR). The second implemented Storvik [19], posterior distribution estimated have supported low-dimensional spaces, sufficient statistics from sample filtered considered. third (PLS) Carvalho’s Particle Learning Smoothing [31]. cash future delivery dates results indicate postponement payment, different maturity highly correlated. Likewise, date last periods year 2017, lower than expiration 12 24 months found, opposite occurs 1 6 months.