RMB CPI Nowcasting Challenge
'%3e%3cpath fill='%23001489' d='M0 0v6h9V0z'/%3e%3cpath fill='%23e03c31' d='M0 0v3h9V0z'/%3e%3cg stroke='%23fff' stroke-width='2'%3e%3cpath id='d' d='m0 0 4.5 3L0 6m4.5-3H9'/%3e%3cuse xlink:href='%23b' stroke='%23ffb81c' clip-path='url(%23c)'/%3e%3c/g%3e%3cuse xlink:href='%23d' fill='none' stroke='%23007749' stroke-width='1.2'/%3e%3c/g%3e%3c/svg%3e)
Helping South Africa
$17 000 USD
Completed (over 2 years ago)
Forecast
1277 joined
21 active
Start
Sep 01, 23
Close
Dec 12, 23
Reveal
Dec 14, 23
September CPI
Data · 18 Oct 2023, 09:45 · 5
headline CPI 111.8
food and non-alcoholic beverages 119.2
alcoholic beverages and tobacco 111.1
clothing and footwear 104.8
housing and utilities 109.4
household contents and services 108.8
health 110.8
transport 116.7
communication 99.5
recreation and culture 106.3
education 110.4
restaurants and hotels 110.7
miscellaneous goods and services 110
not bad, but unfortunately I overestimated "restaurants and hotels" too much
Well done, Ihar!
It appears that "restaurants and hotels" posed a challenge for a few of us. However, the most significant deviation for my model was in the "transport" sub-category, exceeding 3 points! What a volatility this was!
I'm curious to know how you managed to manage this beast—your strategy must be truly remarkable.
Given that we can't modify our models after the initial submission, I believe it would be beneficial for us to exchange insights and approaches.
Thank you, JuliusFx!
There is no "magic" in my strategy. I just use simple ML models from scikit-learn with some kind of hyperparameter tuning and then select the model with the lowest RMSE among them for each sub-category based on the validation with last 24 months of data.
Wow! I am also using some ML models too Linear Regression and Ridge to be specific. I guess they do not adapt well to big surprieses because they are not tuned. Eager to see how we perform in October now.
and again "restaurants and hotels" makes something strange: 110.7 (September) -> 114.2 (October) = +3.2%
In addition, transport also experiences a similar jump as before, rising from 116.7 in September to 120 in October, reflecting a 2.8% increase. The earlier 2.7% rise from 113.6 in August to 116.7 in September took a toll on most of us.
Very tough!