The probability of fire occurrence ‘Pf’ should be function of so many variables including:
1 - 'P1' percentage related to previous statistics and it should be at least multiplied by a coefficient of 1.5, since the real world statistics are far more serious than the one recorded and compiled by statisticians... etc
2 - 'P2' the percentage of availability of combustible material in the building compared to other none combustible ones... etc
3 - 'P3' the percentage related to the medium temperature within the building deducted from a year time records... etc
Probability of fire occurence ‘Pf’ = f (P1, P2, P3, ...) = P1 + P2 + P3 + ... etc
In addition to other variables where applicable... etc, also each sub probability could be function of others sub-variables… etc
What do you think guys?