I am training a model for crop yield prediction having a self-constructed dataset of 6 features and 2000 records. However, the dataset is biased and I am not getting accurate results. I have tried different algorithms but the same output. Did some pre-processing but still the same results.
Here is the co-relation matrix:
Correlation of features with the target variable (production(kg)):

• Region: NaN
• CropType: NaN
• Year: -0.019087
• FieldArea(acres): 0.925150
• Temperature(dc): 0.020016
• WaterAvailablity(m3): -0.049801
• SoilType: 0.028538
• Name: production(kg), dtype: float64

Please suggest me some solution for this

Want the solution of biasness