Lewis-Gitari/Tanzania-water-wells

Tanzania, as a developing country, struggles with providing clean water to its population of over 57,000,000. There are many water points already established in the country, but some are in need of repair while others have failed altogether. Build a classifier to predict the condition of a water well. # TANZANIA WATER WELLS ANALYSIS # OVERVIEW The Tanzanian government, foreign donor organizations, churches and faith-based organizations, and even the villagers have all been involved in funding and installation of wells. Some of the installed wells however cease to function over time while others remain in dilapidated conditions needing repair. For this project i'll be working as a Data scientist along an NGO alligned towards provision of water among communities in Tanzania. I'm to analyse existing data within the Tanzanian water ministry and create a model that will guide us on the functionality estimated life span of wells in different regions and varying conditions. # TOOLS Python Jupyter Notebook #### Python Libraries used Pandas: Data pre_processing and cleaning Matplotlib: Data visualization Seaborn: Data visualization Numpy: Data pre_processing #### ML ALGORITHM used Scikit_learn: Modelling # WORKFLOW ## 1. Data preparation + Imported the necessary packages + Loading the training-set-values dataset as downloaded + Checked for the general data shape + Concatenating the training and testing data sets. + Checking for duplicates in the dataset. + Checking and handling missing values in our dataset. + Feature engineering. + Dropping columns that are of low impact to model outcome. ## 2. Data analysis & visualization Conducted concurrent analysis and visualization, i'm trying to show the relationship between different columns in our dataset.E.g relationship between the ['funder'& 'region']. Displayed the relationship between the numerical data E.g ['amount_tsh', 'gps_height', 'population', 'age_of_well'] github.com ## 3. Modelling We'll first identify the categorical and numerical columns for ease of one hot encoding to allow easy fitting to different models. + categorical data _columns = ['funder','basin','region', 'public_meeting', 'permit', 'extraction_type_group','management','payment_type','water_ …