100 days of AI/ML. Study or apply AI/ML for at least 1 hour each of the next 100 days, starting at July 23.
Work on a Random Forest model to predict urban fire incidents in Hermosillo, México using Servicio de Bomberos de Hermosillo open database
Worked on Random Forest model for Saturdays.ai for prediction of fire incidents in Hermosillo. Played with MLaaS software Intersect Labs, the software is very limited since it will not allow you to do any data preprocessing and will not let you work with categorical data before their algorithm starts. I am building a MLaaS myself for lead scoring.
Worked on cleaning a dataset for one of our customers that will be used for predicting churn rate. I also made updates to this week's python code for Saturdays.ai Hermosillo class.
Worked on Random Forest prediction model for Saturdays.AI class. Prepared material for Intro to Random Forests - Favorita Grocery Sales Forecasting Kaggle Challenge that will be share with the class. We are expecting to have a blog post for the Fire Prediction after the groups' presentations.
Worked on Random Forest model with the Saturdays.AI class. Teached them the basics for description analysis, feature enginerring and model creation.
Worked on a new notebook for the Hermosillo Fire Deparment dataset for incident predictions. Obtained the following results:
precision recall f1-score support
ATENCION PRE HOSPITALARIA 0.50 0.58 0.54 100
COBERTURA DE SEGURIDAD 0.41 0.33 0.36 172
DERRAME DE COMBUSTIBLE 0.30 0.33 0.31 46
ENJAMBRE DE ABEJAS 0.93 0.98 0.95 604
FALSA ALARMA 0.65 0.65 0.65 207
FUGA DE GAS LP O GAS NATURAL 0.74 0.81 0.78 229
INCENDIO DE CASA DE CARTÓN 0.50 0.16 0.24 82
INCENDIO DE CASA DE MATERIAL 0.47 0.50 0.48 288
INCENDIO DE COMERCIO 0.52 0.29 0.37 45
INCENDIO DE INSTALACIÓN PÚBLICA 0.00 0.00 0.00 22
INCENDIO DE VEHÍCULO 0.77 0.87 0.82 153
INCENDIO DE ZACATE Y BASURA 0.88 0.90 0.89 1556
RESCATE 0.38 0.31 0.34 65
RETIRO DE ANIMAL 0.73 0.75 0.74 107
accuracy 0.77 3676
macro avg 0.56 0.53 0.53 3676
weighted avg 0.76 0.77 0.76 3676