IEOR 165 Homework 5 Due April 30, 2015 Question 1. a. Assuming the weather on a given day only depends on the weather on the previous day, weather forecasting can be modeled as a Markov Chain. Weather can be sunny, rainy or cloudy. The data coming from the US Weather Service can be summarized as follows: today sunny today rainy today cloudy tomorrow sunny 286 128 245 tomorrow rainy 123 333 255 tomorrow cloudy 165 256 189 Find the maximum likelihood estimates of the transition probabilities. b. Each time a certain horse runs in a three-horse race, he can win, come in second, and come in third. This race is modeled as Markov chain and the following data is collected. first second third first 55 44 20 second 71 62 39 third 22 56 53 Find the maximum likelihood estimates of the transition probabilities. Question 2. A popular brand of tennis shoe has had the following demand history by quarters over a three-year period. 1990 1 2 3 4 Demand 12 25 76 52 1991 1 2 3 4 Demand 16 32 71 62 1992 1 2 3 4 Demand 14 45 84 47 a. Using the data from 1991 and 1992, determine initial values of the intercept, slope and seasonal factors for Winter’s method. 1 b. Assume that the observed demand for the first quarter of 1993 was 18. Using α = 0.2, β = 0.15 and γ = 0.1, update the estimates of the series, the slope and the seasonal factors. c. What are the forecasts made at the end of the first quarter of 1993 for the remaining three quarters of 1993? Question 3. The energy consumption can be modeled as a the linear time invariance (LTI) dynamical system in discrete time. xt+1 = Axt + But The following data shows how energy consumption (thousand metric ton of oil equivalent per capita) is related to the Gross Domestic Product (GDP) per capita. Write a code to determine the values of A and B. GDP per capita Energy Consumption (thousand metric ton of oil equivalent per capita) 16,176 5468.54 16,472 5566.28 15,779 5665.80 16,076 5764.13 16,836 5864.23 17,582 5967.52 17,862 6073.97 18,348 6182.22 18,993 6292.91 19,108 6406.52 18,872 6521.51 18,255 6636.81 18,139 6751.31 18,303 6866.49 18,936 6983.20 19,227 7102.85 19,319 7224.46 19,953 7347.44 20,613 7473.42 21,563 7602.52 22,488 7735.64 22,599 7872.72 23,156 8011.37 23,409 8152.94 23,943 8296.55 24,470 8443.04 24,971 8592.44 25,809 8744.69 25,267 8900.81 2 Solution 1. a. p11 = p12 = p13 = p21 = p22 = p23 = p31 = p32 = p33 = n11 n11 + n12 + n13 n12 n11 + n12 + n13 n13 n11 + n12 + n13 n21 n21 + n22 + n23 n22 n21 + n22 + n23 n23 n21 + n22 + n23 n31 n31 + n32 + n33 n32 n31 + n32 + n33 n33 n31 + n32 + n33 = = = = = = = = = 286 286 + 123 + 165 123 286 + 123 + 165 165 286 + 123 + 165 128 128 + 333 + 256 333 128 + 333 + 256 256 128 + 333 + 256 245 245 + 255 + 189 255 245 + 255 + 189 189 245 + 255 + 189 = = = = = = = = = 286 574 123 574 165 574 128 717 333 717 256 717 245 689 255 689 189 689 = 0.498 = 0.214 = 0.288 = 0.179 = 0.464 = 0.357 = 0.356 = 0.370 = 0.274 b. p11 = p12 = p13 = p21 = p22 = p23 = p31 = p32 = p33 = n11 n11 + n12 + n13 n12 n11 + n12 + n13 n13 n11 + n12 + n13 n21 n21 + n22 + n23 n22 n21 + n22 + n23 n23 n21 + n22 + n23 n31 n31 + n32 + n33 n32 n31 + n32 + n33 n33 n31 + n32 + n33 = = = = = = = = = 3 55 55 + 71 + 22 71 55 + 71 + 22 22 55 + 71 + 22 44 44 + 62 + 56 62 44 + 62 + 56 56 44 + 62 + 56 20 20 + 39 + 53 39 20 + 39 + 53 53 20 + 39 + 53 = = = = = = = = = 55 148 71 148 22 148 44 162 62 162 56 162 20 112 39 112 53 112 = 0.371 = 0.48 = 0.149 = 0.272 = 0.383 = 0.346 = 0.179 = 0.348 = 0.473 Solution 2. a. Seasonal factor initialization Dt Vi − [(N + 1)/2 − j]G0 where i = 1 for the first season and i = 2 for the second season and j is the period of the season. Initialization procedure: ct = V1 1 = N V2 = G0 = S0 = c−7 = c−6 = c−5 = c−4 = c−3 = c−2 = c−1 = c0 = c−3 c−2 c−1 c0 sum: 1 N −N ∑ 1 Dj = (16 + 32 + 71 + 62) = 45.25 4 j=−2N +1 0 ∑ 1 Dj = (14 + 45 + 84 + 47) = 47.5 4 j=−N +1 V2 − V1 47.5 − 45.25 = = 0.5625 N 4 3 N −1 = 47.5 + (0.5625) = 48.34 V2 + G0 2 2 16 = 0.36 45.25 − [5/2 − 1]0.5625 32 = 0.711 45.25 − [5/2 − 2]0.5625 71 = 1.559 45.25 − [5/2 − 3]0.5625 62 = 0.345 45.25 − [5/2 − 4]0.5625 14 = 0.30 47.5 − [5/2 − 1]0.5625 45 = 0.953 47.5 − [5/2 − 2]0.5625 84 = 1.758 47.5 − [5/2 − 3]0.5625 47 = 0.972 47.5 − [5/2 − 4]0.5625 avg seasonal factors (c−7 + c−3 )/2 = 0.33 (c−6 + c−2 )/2 = 0.832 (c−5 + c−1 )/2 = 1.659 (c−4 + c0 )/2 = 1.159 3.98 4 normalized seasonal factors 0.331 0.837 1.667 1.165 4 b. α = 0.2 β = 0.15 γ = 0.1 D1 = 18 S1 = α(D1 /c−3 ) + (1 − α)(S0 + G0 ) = 0.2(18/0.33) + 0.8(48.34 + 0.5625) = 49.97 G1 = β(S1 − S0 ) + (1 − β)G0 = 0.15(49.97 − 48.34) + 0.85(0.5625) = 0.72 c1 = γ(D1 /S1 ) + (1 − γ)c−3 = 0.1(18/49.97) + 0.9(0.331) = .3345 c. At this point it is recommended to renorm c−2 , c−1 , c0 and the new value of c1 to add to 4. c1 c−2 c−1 c0 sum: avg seasonal factors 0.334 0.837 1.667 1.165 4.003 normalized seasonal factors 0.334 0.836 1.666 1.164 4 Forecasts for 2nd , 3rd and 4th quarters of 1993 F1,2 = [S1 + G1 ]c−2 = (49.97 + .72)0.836 = 42.378 F1,3 = [S1 + 2G1 ]c−1 = (49.97 + 2(.72))1.666 = 85.66 F1,4 = [S1 + 3G1 ]c0 = (49.97 + 3(.72))1.164 = 60.67 Solution 3. A = 1.009 B = 0.003 5