WEBVTT 00:01.710 --> 00:08.740 OK, so let us first discuss the solution for the first project that this house price prediction problem. 00:09.330 --> 00:17.860 So here is what we have done is we have imported all the required libraries and we have got the data. 00:18.270 --> 00:24.690 So here we have the data said that this house data set and this data set contains around twenty one 00:24.690 --> 00:30.360 thousand rooms and twenty one columns out of these. 00:30.720 --> 00:33.300 We have selected these different data. 00:33.300 --> 00:33.790 Right. 00:34.320 --> 00:39.060 So we have got the database, which includes object. 00:39.330 --> 00:41.640 So we have got this data. 00:42.330 --> 00:50.940 And next, what we are doing here is checking that we have only this data column, which is a timestamp. 00:50.940 --> 00:53.850 So we will basically ignore this data column. 00:54.360 --> 01:02.400 So what we are doing here is we are checking if there are any columns which are having not a numbers 01:02.430 --> 01:03.030 of value. 01:03.390 --> 01:11.070 And then we see that we get that there are no particular columns, which are no particular rules, which 01:11.070 --> 01:14.190 have a number in the values. 01:15.360 --> 01:22.790 So we see that the data is pretty much structured and it does not have any norten no values. 01:23.190 --> 01:29.780 So we can basically jump in to finding the correlation between the features and data equity. 01:29.810 --> 01:38.280 But so now when we are finding out the correlations, so we have simply got the feature values and the 01:38.280 --> 01:43.230 target values and we are finding out the correlation. 01:43.260 --> 01:50.100 So this is the blind correlation list, which we have created and we are putting in the Peerson correlation 01:50.100 --> 01:57.780 inside this, using this bias in our function and comparing each of those values with the target value. 01:57.810 --> 02:04.470 So we are comparing the all the features with the target value to find out the correlation values. 02:05.670 --> 02:12.360 So next, what we are doing is we are converting all this data, which we have got for the correlation 02:12.360 --> 02:13.890 in toward the Gulf Stream. 02:14.310 --> 02:24.210 And we have created this data correlations, data frame for that and for each location they are checking 02:24.240 --> 02:31.380 and they are getting the values, the absolute values, and we are sorting based on these correlation 02:31.380 --> 02:31.980 values. 02:33.270 --> 02:43.290 So here we can see the maximum correlation value of which we have here is you can see that here it is, 02:43.290 --> 02:53.610 70 percent correlation and it goes down to zero point zero two, which is basically two percent correlation 02:53.610 --> 02:54.590 between the values. 02:54.840 --> 03:00.840 So you can see the minimum correlation that we have is between long and prices and conditions and prices, 03:01.080 --> 03:07.500 while the maximum correlation which we have is of seventy five, seventy percent between square feet 03:07.500 --> 03:08.790 living and the price. 03:10.740 --> 03:17.000 So we can see that the top five features are most correlated features with the target price. 03:17.280 --> 03:23.220 And so what we do is we applaud the best tool aggressor's jointly. 03:23.220 --> 03:25.440 So we applaud this. 03:25.470 --> 03:26.970 So how do we do that? 03:27.300 --> 03:36.270 So we are taking this very column and this X column this week column would consist of square feet living 03:36.300 --> 03:42.180 on the three, which are the top two columns which have the maximum correlation. 03:44.040 --> 03:52.710 And we are again sorting the values so that we can see a proper plot with these values and we are creating 03:53.040 --> 03:55.650 this y value y es shape. 03:55.650 --> 04:00.110 So we are getting the values with respect to that index. 04:02.310 --> 04:07.380 So we are getting the X values and Y values corresponding to these next. 04:07.380 --> 04:11.280 We are generating the different plots for that. 04:11.280 --> 04:16.720 So we are generating four square feet with grade and prices. 04:16.740 --> 04:18.520 So here we have the data. 04:18.540 --> 04:20.500 So this is the square data. 04:20.520 --> 04:23.730 This is the three data and this is the price data. 04:24.210 --> 04:30.980 So here you can see how these plots are very much similar to each other. 04:31.410 --> 04:39.030 And you can see that these square feet and great plots are almost similar, only difference being the 04:39.420 --> 04:42.100 range and the scale a little bit. 04:42.360 --> 04:45.540 Otherwise, everything else seems to be a lot similar. 04:46.890 --> 04:52.440 So next, what we will be doing is we will fit the model first. 04:52.590 --> 04:57.360 So we are generating the base model, using the menials regression. 04:57.630 --> 04:59.700 So when we are generating this base. 05:00.300 --> 05:09.230 We are getting some of the top forelimbs that is you can see here we have square feet living green, 05:09.240 --> 05:14.620 square feet above square feet, living bathroom to your bedroom. 05:14.640 --> 05:16.080 So let's see here. 05:18.120 --> 05:23.180 We have these floors, all of these. 05:24.030 --> 05:27.080 So these are different columns which we are taking off. 05:28.590 --> 05:39.530 So now we are getting the X and Y columns and the performing cross-validation on this, trying to split. 05:39.540 --> 05:42.010 So we are getting the split for this data. 05:42.030 --> 05:49.320 We are simply getting the split of the data, which is the size of zero point two. 05:49.320 --> 05:56.130 That is, 20 percent of data will belong to the best data and the rest will be a part of our training 05:56.130 --> 05:56.760 dataset. 05:58.140 --> 06:01.920 So here we have extreme X test wiping rightest. 06:02.610 --> 06:11.520 So next, what we have done is we have this regression and then we have predicted the values and we 06:11.520 --> 06:16.620 are scoring this regression and the score gives us zero point seven zero. 06:18.600 --> 06:27.090 So here we are getting around 70 percent prediction score, which is not a very good score and not a 06:27.090 --> 06:28.200 very bad score. 06:28.230 --> 06:35.670 This is a decent point, but yes, because we have implemented the linear model, the line is a Christian, 06:36.090 --> 06:40.080 so it can be improved a lot more for thought. 06:40.110 --> 06:41.460 So let's go ahead. 06:42.900 --> 06:49.860 So we are calculating the square area for this, which comes out to be one nine three six one five. 06:50.130 --> 06:57.000 Now, one thing to note here is that the root mean squared error is the squared value squared value 06:57.000 --> 07:03.670 error and house prices are again amounts which are larger value. 07:03.690 --> 07:10.080 That is why the item is he is also a large of what you are asking is to bring it closer to zero. 07:10.410 --> 07:14.460 The value is large because it is a house price. 07:14.460 --> 07:17.800 If it was an interest rate, would have been a very small value. 07:18.150 --> 07:22.700 So there is nothing to be scared about out of this particular value. 07:22.710 --> 07:23.730 It's just the scale. 07:24.930 --> 07:27.960 So next, what we are doing is we are implementing exhibits. 07:28.920 --> 07:35.190 So here I have taken up an estimated one hundred and eighty two point zero eight and I have three in 07:35.550 --> 07:37.080 this particular model. 07:37.110 --> 07:45.540 And after cleaning I get the ingredients code as eighty one point three seven percent, which is almost 07:45.540 --> 07:48.660 close to eighty five percent. 07:48.660 --> 07:54.810 Eighty four percent of what we can do is this is just a simple implementation. 07:54.810 --> 08:01.560 Using examples, suppressing what you will be doing is you can similarly implement a random forest, 08:01.560 --> 08:10.140 extremely distressed, or you could implement the different algorithms got stuck together and these 08:10.140 --> 08:13.280 could be I could achieve a lot more better result. 08:13.800 --> 08:19.070 So that is something which you will have to do and you will have to implement more modules. 08:19.080 --> 08:24.120 You will have to compare different models and then get better results out of the.