WEBVTT 00:02.040 --> 00:09.810 In this session, we will discuss about dissolution of the used car price prediction problem, so this 00:09.810 --> 00:15.750 Craigslist, which is the data center, which we have, is the world's largest collection of huge vehicles 00:15.750 --> 00:16.530 for sales. 00:16.890 --> 00:24.780 And this particular dataset includes every used vehicle for within the United States, which is present 00:24.780 --> 00:25.570 on Craigslist. 00:26.370 --> 00:33.640 So initially we would include the libraries, which are known by abundance, even as I live in invalidly 00:33.640 --> 00:34.930 leadership and were of off. 00:36.240 --> 00:43.380 So as you can see, this particularly date Kansas, the I.D., UARS region, region area and many other 00:43.380 --> 00:46.880 object type data columns. 00:47.130 --> 00:54.810 So we will be converting these data columns into medievalists and such data columns which don't really 00:54.810 --> 00:58.950 have any relevance related to the price we would be dropping them. 01:00.840 --> 01:06.680 So the dataset contains around five like thirty nine thousand rolls of data. 01:07.350 --> 01:15.600 And these are the columns which are you as you can see, the columns you are in the region, you all 01:17.280 --> 01:22.150 image you are in these columns have no relevance to us. 01:22.170 --> 01:31.110 So we would be removing these columns and the other object types we will be analyzing and then seeing 01:31.110 --> 01:35.360 if we will have to convert them into the dummy beads. 01:37.860 --> 01:42.520 So the first thing which we can do is we can have a look at the value columns. 01:42.810 --> 01:51.030 So if you see the Idee column has one as the value count, that means that these all are unique values. 01:51.330 --> 01:53.670 So we will be getting rid of this one. 01:54.270 --> 02:00.270 Similarly, the yuan is again unique so we can get rid of them. 02:00.990 --> 02:02.500 Next comes the region. 02:02.550 --> 02:06.350 So there are a total of four hundred and three regions. 02:06.720 --> 02:15.780 So we will have to find out a few regions which we need to keep and get rid of all others. 02:16.080 --> 02:24.690 So if you see the top five regions can be a high number of values, that is around four thousand, three 02:24.690 --> 02:31.410 thousand and the ones below the earth towards the bottom, they can be around thirty eight or thirty 02:31.410 --> 02:32.910 seven guys and their listings. 02:34.020 --> 02:36.670 So we will be selecting only the top of him. 02:38.070 --> 02:44.850 Next comes the region guaran so again, we will be getting rid of these region bureaus because we cannot 02:45.150 --> 02:50.480 delay much from this next if we look at the price. 02:50.730 --> 02:59.010 So you can see that for zero as a price, we have around forty four thousand barrels of data, which 02:59.250 --> 03:01.020 cannot really be possible. 03:01.410 --> 03:09.500 And towards the bottom, again, a lot of cars have a very low price. 03:10.530 --> 03:14.870 So there are certain prices which have only one values available. 03:15.150 --> 03:20.330 But again, these are not of our problem because these are integer type. 03:20.580 --> 03:28.290 So the only thing which we need to worry about here is this particular dataset where we have zero as 03:28.290 --> 03:30.180 the value of the price. 03:30.420 --> 03:32.440 So that is something which is not possible. 03:32.460 --> 03:34.240 So this is something which we need to be careful. 03:35.400 --> 03:36.950 So why are you taking care of this? 03:37.290 --> 03:47.010 What you can do is you can be a mean value or a median value of that particular region where it belongs. 03:47.280 --> 03:55.110 Or you can drive on other different methods or find different patterns between the regions or the type 03:55.110 --> 04:01.600 of cars and then find out the median or mean value accordingly and then put that here. 04:02.820 --> 04:06.530 So that is completely up to you, how you want to use these values. 04:06.960 --> 04:11.330 So because this is a large number you can automatically take on me. 04:12.630 --> 04:20.400 So I would suggest you to take a mean of different regions and then infuse those values in these blank 04:20.400 --> 04:20.850 values. 04:23.610 --> 04:25.210 Next, we have here. 04:26.460 --> 04:30.710 So these are the years when these cars were manufactured. 04:31.020 --> 04:36.900 So here you can see that there are a few cars which are very old. 04:37.170 --> 04:42.860 And most of and here we have a lot of cars which are buried under recent years. 04:42.870 --> 04:50.070 So what you can do is for the cars, which have very old data, you can basically create a separate 04:50.070 --> 04:58.650 column or maybe see older than 50 years or older than 40 years and create a single category out of it 04:58.860 --> 04:59.850 instead of having. 05:00.130 --> 05:01.060 These combined. 05:03.410 --> 05:13.160 Similarly, here we have some of the fringe manufacturers of these cars, so you can see that there 05:13.160 --> 05:19.790 are a few manufacturers which have a lot of cars listed here, while some manufacturers have a very 05:19.790 --> 05:24.900 few number of cars listed like the Abdeslam, Harley Davidson, Alfa Romeo. 05:25.070 --> 05:28.680 So what you can do is you can put these into other categories. 05:28.680 --> 05:34.100 Only of these categories don't really make a huge impact on the prices. 05:34.640 --> 05:38.480 But you can see that this actually makes an impact. 05:38.480 --> 05:45.670 Like if we have Ferrari, Aston Martin, these cars are actually having a very high price. 05:45.950 --> 05:49.220 So this will actually have a huge impact on the prices. 05:49.430 --> 05:56.540 So we cannot really converted directly without having a good analysis on this. 05:56.660 --> 06:01.430 So we will find out and divide these into different categories. 06:01.430 --> 06:08.030 Or you can divide these manufacturers into different levels which are associated with the price. 06:08.960 --> 06:10.910 So this is another thing that you can do. 06:15.870 --> 06:26.140 Next, we have the models now, these models are actually very unique things, and these are very specific 06:26.140 --> 06:28.350 needs for a particular God. 06:28.350 --> 06:36.510 I'm here if you see here we have this Silverado fifteen hundred and this name has been split inside. 06:36.750 --> 06:42.740 And I have another six thousand rows of the David fifteen hundred and another 5000 bottles of the Dovid 06:42.750 --> 06:43.460 Silverado. 06:43.830 --> 06:50.100 So which means that these are kind of replicative and it does not really have a good information present 06:50.100 --> 06:56.030 here because the same guy has been listed in three different categories inside this. 06:56.310 --> 07:03.720 So what we can do is we can actually decide if we need to keep this model or not or if we need to convert 07:03.720 --> 07:07.890 this into a different type of category or how we need to deal with it. 07:07.920 --> 07:09.870 This is something that we will have to find. 07:14.570 --> 07:23.210 Next are the conditions, so because these are very lesser numbers, we can have all these conditions 07:23.210 --> 07:33.050 and read them and convert them into an ordinary variable, which looks like the best way out, then 07:33.050 --> 07:38.160 we have the number of cylinders which can directly be converted into numeric value. 07:40.070 --> 07:41.510 Then we have FuelCell. 07:41.510 --> 07:49.940 These can be converted into a categorical dummy variable and then we have all the which is statically 07:49.940 --> 07:51.380 an integer type. 07:51.380 --> 07:54.890 So we don't really need to think much about it. 07:57.500 --> 07:59.680 Then comes the title. 08:00.330 --> 08:03.700 So this needs to be converted into dummy variables. 08:04.100 --> 08:07.150 Then we have the transmission again. 08:07.160 --> 08:09.530 This would be converted into dummy variables. 08:10.310 --> 08:19.670 After that, we have a van, which is basically the details about the car, which seem to be a lot unique, 08:19.670 --> 08:25.470 as there are around one like eighty one thousand different values associated with it. 08:25.640 --> 08:29.300 So the best way out would be getting rid of this particular column. 08:32.190 --> 08:38.040 Next, we have the type of drive which can be directly converted into a dummy, the. 08:40.520 --> 08:49.040 Size can again be converted into the type of the car can again be converted into a dummy variable, 08:50.600 --> 08:51.960 then paint color. 08:51.990 --> 08:59.690 We can maybe decide the top four colors and based on the top four colors, we can see if we need to 08:59.690 --> 09:01.190 convert them into three variables. 09:01.190 --> 09:06.410 And we can actually get rid of the view towards the bottom because they are a lot of colors. 09:06.410 --> 09:13.820 We don't want to create a lot of columns out of just colors because that would hardly have any impact 09:13.820 --> 09:14.540 on the pricing. 09:16.250 --> 09:19.210 Next, we have image euro, which we can directly get rid of. 09:20.630 --> 09:25.620 After that, we have this huge text column, which is description. 09:26.030 --> 09:28.550 So there are two ways. 09:28.920 --> 09:35.120 First would be you can create a model by getting rid of this column entirely. 09:35.430 --> 09:44.900 And the second would be you can write the first method and additionally create counterfeiter's or Defitelio 09:44.900 --> 09:51.500 vectors out of this description column and add that as an additional feature to your mom. 09:54.090 --> 10:02.940 Next comes the state, and if we look at the state, there are a lot of values in the state, so we 10:02.940 --> 10:09.110 will have to kind of get rid of this column because it does not have much information in it. 10:09.690 --> 10:14.970 So we will have to find out if there is a specific pattern which is associated regarding priorities 10:14.970 --> 10:15.730 for the state. 10:15.960 --> 10:18.390 If not, then we can radically get rid of this. 10:20.270 --> 10:25.790 Next comes the latitude and longitude, which we can directly get rid of, because that would again 10:25.790 --> 10:27.170 be similar to the state. 10:31.070 --> 10:36.830 So what we are doing here is we are dropping the idea you are in the region you are living in means 10:36.830 --> 10:43.390 you are description, latitude, longitude, country, region, and this is something which I have done. 10:43.580 --> 10:49.940 So I also told you about what all you can do other than what I am actually implementing here so you 10:49.940 --> 10:51.380 can do it accordingly. 10:51.390 --> 10:55.030 But this is just my implementation. 10:57.880 --> 11:05.290 Next, this is the final list of columns that we have, so we have praise your manufacturer, modern 11:05.680 --> 11:14.560 condition, cylinder fuel, odometer titles, distance mission, drive, size type being color and speed. 11:17.030 --> 11:20.540 Now, let's describe this and then we see the. 11:21.050 --> 11:29.060 So we have praise so you can see the minimum price is zero and then we have twenty five percent of the 11:29.060 --> 11:31.650 values which are price around four. 11:32.150 --> 11:37.480 Then we have eer of manufacturing. 11:38.750 --> 11:41.600 After that we have the odometer values. 11:41.870 --> 11:45.020 So first thing that we going be doing is handling outliers. 11:45.380 --> 11:53.000 Now, as you have already seen, that the prices have a lot of outlier present because there are a lot 11:53.000 --> 11:55.340 of old cars which are also present. 11:55.910 --> 12:04.040 So what we will be doing is we will see the find out the outlier of the will for the standard variable 12:04.040 --> 12:04.520 itself. 12:05.270 --> 12:07.700 And the target variable here is price. 12:07.710 --> 12:08.010 Right. 12:08.030 --> 12:17.420 So we will be finding out the outliers that and we will remove the model which so that those particular 12:17.420 --> 12:24.110 values could be accurate because these outliers present in the price itself, that is the target itself, 12:24.440 --> 12:31.460 will bring in a lot of difference in the values of the mean and these deviation which we have in the 12:31.460 --> 12:31.940 values. 12:32.930 --> 12:36.680 So the first thing which we will be doing is removing outliers from the licensing. 12:37.550 --> 12:42.490 So the difference between the 75 percent value and the maximum value is very large. 12:42.710 --> 12:47.490 So we will leave the 10 percent values at the end of the distributions. 12:47.810 --> 12:53.690 So what we are doing is we are finding out on day one what they do and then finding out the percentile 12:53.690 --> 12:54.350 values. 12:54.530 --> 13:03.260 And we will be dropping all the rules of the the list and these two values, which we have just found 13:03.260 --> 13:03.500 out. 13:06.760 --> 13:09.950 So now we will have a look and he ordered me to call you. 13:10.340 --> 13:14.480 So this is a distorted values of the order values. 13:15.850 --> 13:23.020 So here you can see the full value is zero and next comes to the eight thousand, which means that there 13:23.020 --> 13:29.340 is a lot of difference between these values and there are a lot of not on numbers as well. 13:36.470 --> 13:43.250 These are the different values which we have, so there are not a number of values and only one zero 13:43.250 --> 13:43.650 value. 13:43.910 --> 13:49.100 So what we will be doing is we will find out of whatever is having not a number. 13:49.430 --> 13:58.310 So here we have around 80000 values which are having a land value, and we will create the scatterplot 13:58.310 --> 13:59.360 for all the leaders. 13:59.380 --> 14:08.050 So here you can see there is one outlier value and all of the values are actually scattered here near 14:08.060 --> 14:08.540 zero. 14:14.380 --> 14:17.050 So this is the maximum value which we have. 14:19.510 --> 14:29.320 So what we are doing here is we will simply find out the values, we will drop this maximum value and 14:29.320 --> 14:33.030 we will drop the column, the particular rule which has the value. 14:34.450 --> 14:38.160 So we will basically be getting rid of this particular data point. 14:38.530 --> 14:44.080 I'm the first data point which have value zero percent in this. 14:47.530 --> 14:50.660 So now when us pulled us out of it. 14:50.890 --> 14:53.590 So this is how the distribution comes out to me. 14:54.760 --> 14:58.240 So regarding the values which have not. 14:58.280 --> 14:58.780 No. 14:59.140 --> 15:05.780 So for them, there are around above three hundred thousand. 15:05.800 --> 15:11.180 These values can be considered as outliers because most of the values are present here. 15:12.160 --> 15:16.220 These are the most of the values and these are the other values which are present here. 15:16.450 --> 15:25.360 So we are considering the values above three hundred thousand as the outlier values and then we again 15:25.370 --> 15:26.510 create the scatterplot. 15:26.530 --> 15:28.580 So here you can see the distribution again. 15:34.090 --> 15:41.710 So this is for more analysis, which has been done on the order with the value being done the year as 15:41.710 --> 15:41.920 well. 15:42.340 --> 15:49.630 So here we have the value of here and here we have the value of price. 15:51.030 --> 15:59.560 So with the distribution of the price, you can see that there is not much relevance of year and price 15:59.560 --> 16:06.380 in the lead up because it is kind of like of uniformly distributed price and your distribution. 16:06.640 --> 16:14.440 So if you see here, we have the year nineteen forty below, which the prices are actually not really 16:14.440 --> 16:18.430 much and there are a few, only a few datapoint also present. 16:18.640 --> 16:20.420 So this looks like an outlier. 16:20.440 --> 16:23.650 Otherwise all the data looks like uniformly distributed. 16:23.920 --> 16:31.450 So what we will be doing is we will be getting rid of these particular rules of the year of manufacturing 16:31.450 --> 16:34.540 is less than 19 for 16:38.140 --> 16:38.540 next. 16:38.810 --> 16:43.040 What we are doing is we will handle the other volumes which are present here. 16:43.060 --> 16:48.790 So I am simply finding out the percentage of null values in each claim. 16:49.510 --> 16:56.500 So here I'm just finding out the number of null values and I'm checking of which columns are present 16:56.500 --> 16:56.830 then. 16:57.190 --> 17:00.580 And I have simply found out the percentage. 17:00.610 --> 17:10.420 So here we have the column name the Josiah's Conditions Cylinder Bangalor Drive, a manufacturing model 17:10.420 --> 17:14.890 transmission's you legal status, then null values, the number of values which are present, which 17:14.890 --> 17:17.680 is the value and the percentage value. 17:18.070 --> 17:25.530 So here, if you see that this is column has it on 64 percent of values, which are actually not antico 17:25.540 --> 17:32.370 column condition has five percent values, which the lender has 33 percent values, which are Benkler 17:32.380 --> 17:33.700 has twenty four percent values. 17:33.700 --> 17:40.210 The General Andreev has twenty one percent type, has 18 percent values which are not apart from that, 17:40.210 --> 17:44.140 all other columns have very few number of values which are actually not. 17:46.270 --> 17:52.960 So these are the values column for condition, which is again having a huge number of value counters. 17:53.770 --> 17:55.870 So these are the different value columns. 17:55.870 --> 17:59.000 You can see excellent has a very huge number of presents. 17:59.040 --> 18:05.640 Then we have code, then we have like new, then we have and we have new and salvage. 18:07.570 --> 18:15.490 Now, the missing values in the condition can be found using odometer as mileage affects the condition 18:15.490 --> 18:16.120 of the car. 18:17.080 --> 18:24.430 So what we are trying to do here is finding out the mean value of the odometer reading group by their 18:24.430 --> 18:25.260 conditions. 18:26.830 --> 18:34.000 So here we have the excellent Otomi, the good old Boomin, the like New or Boomin Salvacion, all the 18:34.000 --> 18:35.970 mean and they're or doing so. 18:35.980 --> 18:38.830 These are the mean values which we have just found out. 18:39.100 --> 18:42.040 And these are the values of what, like new. 18:42.040 --> 18:45.120 We have eighty seven thousand for excellent. 18:45.130 --> 18:48.340 We have a little higher value for good. 18:48.340 --> 18:55.900 We have then something, then we have and then we have salvage. 18:56.620 --> 18:59.920 So these are the different mean values for the old. 19:01.960 --> 19:10.060 Now using these values, we can actually include the odometer reading, which will have more value in 19:10.060 --> 19:10.670 the condition. 19:11.410 --> 19:18.770 So what we do here is we are simply checking that you so wherever the air is greater than 2090. 19:19.390 --> 19:30.970 So there for the condition we are basically putting as new, then wherever the value is, make new or 19:30.970 --> 19:34.660 less than make new the mean value of the like new. 19:35.030 --> 19:45.490 We are simply putting as the like new that whatever the value is again greater than fair or don't mean, 19:45.850 --> 19:49.500 then we are putting forward that it is greater than all the mean. 19:49.510 --> 19:55.610 And this is basically a condition which you can actually analyze and find out what has been done here. 19:55.900 --> 19:59.530 So based on that, we have simply rated excellent good in salvage. 20:00.670 --> 20:06.440 Then what we are doing here is we are finding out the values again, the percentage values. 20:06.730 --> 20:11.220 So here again, you can see this is 60 for which we have just seen. 20:11.470 --> 20:15.120 So we are dropping all the values which have less than five percent value. 20:15.170 --> 20:17.650 So here we have this manufacturer. 20:17.920 --> 20:20.590 So below this all comes under the five percent. 20:21.220 --> 20:28.360 So we are removing the null value rules from the title status, human transmission model and manufacture. 20:30.520 --> 20:32.260 Next, we are also dropping the. 20:32.710 --> 20:39.100 With more than 30 personal views, but cylinder's can be an important feature, so we are not dropping 20:39.100 --> 20:40.010 the cylinder for you. 20:40.480 --> 20:46.890 OK, so we have simply dog this size clip from this next Izadi Malverde. 20:48.280 --> 20:50.620 So we are just finding the values again. 20:51.640 --> 20:54.020 So cylinder has 33 percent null value. 20:54.040 --> 21:00.090 We have been color drive by the null values now and on the column you don't really have any values. 21:01.450 --> 21:09.430 So we are finding out the paint color and we have simply filling them up using the full method. 21:09.850 --> 21:15.620 Similarly, we are finding out the values for drive and filling it using the Hollywood film method and 21:15.640 --> 21:17.460 same in place to type and. 21:19.990 --> 21:28.330 This leaves us with a cylinder having 11 values and drive, having two null values which we can actually 21:28.330 --> 21:30.720 get rid of because that is a very small number. 21:34.390 --> 21:42.250 Now, let's explore this later a little bit, and now we're importing sci fi and importing stats from 21:42.250 --> 21:44.590 sci fi and they're creating a lot of. 21:46.690 --> 21:55.660 If you see we have a price and we have here here, so you can see price and you don't really have much 21:56.080 --> 21:57.020 association. 21:57.040 --> 22:00.120 These are actually a uniform distribution present here. 22:03.040 --> 22:07.730 So we can find out these different relationships which are present here. 22:07.750 --> 22:10.590 So here we have the meter and all the window which comes out. 22:10.600 --> 22:15.790 Seeing here and here again would not have much relevance, price and pricing and we wouldn't have much 22:15.790 --> 22:17.140 relevance here. 22:17.140 --> 22:23.480 And prices are simply distributed by when we talk about all the meter in price. 22:23.710 --> 22:33.150 So as the price of basically as the odometer value increases, the price is also increasing. 22:33.940 --> 22:38.560 So as the volume, the value is increasing, the prices actually. 22:39.620 --> 22:40.250 Decreasing. 22:44.350 --> 22:50.080 So this is the relationship which we have and these are the details which we are live with now, for 22:50.080 --> 22:55.610 all the objects, we will simply be converting them into dummy columns. 22:55.960 --> 23:04.180 So here you can see these are the different state values, which we have a lot in number and different 23:04.180 --> 23:05.560 values associated as well. 23:06.040 --> 23:09.330 So now we are comparing the condition with the price. 23:09.340 --> 23:13.430 So these are different conditions and these are different prices. 23:13.450 --> 23:19.170 So as you can see, for knew the prices high and for fear the price is the least. 23:20.680 --> 23:26.690 So clearly the would condition new has highest price, as one would expect. 23:27.040 --> 23:32.410 And now we will create a categorical plot between the cylinder and the price. 23:32.710 --> 23:35.170 And this is the volume from which we are creating. 23:36.670 --> 23:39.340 So this is the wilin plot. 23:39.730 --> 23:44.740 So this shows the distribution of number of cylinders and the price. 23:47.200 --> 23:54.680 So here you can see this has 12 cylinders, so this is the price, I mean, at 10000. 23:55.360 --> 24:00.250 Then we have eight Lendell, which has the mean at a little greater than 10000. 24:02.890 --> 24:07.750 Next, we are comparing the fuel with price. 24:08.540 --> 24:10.130 So it's here. 24:10.160 --> 24:13.060 You can see if the fuel type is diesel. 24:13.780 --> 24:15.610 The price is a little higher. 24:15.620 --> 24:21.310 When the fuel type is gas, the prices are lower than for hybrid fuel. 24:21.340 --> 24:24.860 Again, the prices are a little lower. 24:26.320 --> 24:33.850 So this shows that the price range would be for a majority of each type of guy is based on the fuel 24:33.850 --> 24:34.270 itself. 24:34.930 --> 24:46.460 So for gas, it is between five to 17 K, for diesel it is from 12 to 20, K for hybrid. 24:46.480 --> 24:48.900 It is from seven to 15. 24:50.080 --> 24:54.830 And for hybrid, it is from seven to 15. 24:54.830 --> 24:57.850 K under it is 11 to 20. 24:59.560 --> 25:01.480 An electric blanket, Waitiki. 25:04.590 --> 25:11.550 So next of you are creating a categorical plot between the Dido's state design, the price, so for 25:11.550 --> 25:14.430 P.M., you can see the price is a little higher. 25:14.460 --> 25:21.510 Well, when we talk about the lean mean that I don't see this as having the highest price meet. 25:22.620 --> 25:27.030 So these are the different distributions for the prices based on the title SIECUS. 25:29.530 --> 25:35.260 Next, we have the distribution among the transmission in place, which clearly shows that the price 25:35.260 --> 25:39.030 for automatic transmission are almost similar. 25:39.040 --> 25:43.750 But when we talk about the other day of the mission, it has a very high price. 25:47.020 --> 25:58.090 Next, we have forward the wars and different types of the drive, so there is not much difference between 25:58.090 --> 25:58.870 both of these. 25:59.230 --> 26:02.450 But the third one, this particular one is a little different. 26:02.470 --> 26:03.810 These are a lot. 26:03.830 --> 26:11.490 See, so what we can see for that is we are distinguishing between the type of vehicle and the price. 26:11.920 --> 26:16.660 So here you can see different type of vehicle, have a lot of different people faces. 26:17.020 --> 26:21.880 The price of truck and pickup is somewhat similar. 26:22.180 --> 26:26.770 When we talk about sedan and the minivan, they are a lot similar. 26:27.010 --> 26:28.780 So these are different type of distribution, which. 26:30.910 --> 26:36.280 So the important observation can be obtained from the above the regarding the price brackets for each 26:36.280 --> 26:37.010 day of the week. 26:37.450 --> 26:44.280 So we can find out the price window for each week from this particular distribution. 26:46.360 --> 26:49.180 And here you see the distribution for colors. 26:49.570 --> 26:57.340 That is not much variation, but still we have a little price distribution, which we can get out of 26:57.340 --> 27:00.860 this based on the quartile values from these. 27:02.920 --> 27:07.690 Then we have a manufacturer which again gives a huge difference between these values. 27:07.960 --> 27:13.930 So we can see manufacturer is actually giving a lot of difference in the prices which we have. 27:14.680 --> 27:21.440 So next, what we are doing is we are creating the models or you can see the dummy variables are prepared. 27:21.490 --> 27:25.210 So we are simply converting the labels into encoded values. 27:25.870 --> 27:30.250 And next, we are simply having a string split of the data. 27:33.840 --> 27:38.450 And we are applying the forest on top of it. 27:39.300 --> 27:46.740 So when we apply the random forest, we are simply finding out the minoxidil then means when error and 27:46.740 --> 27:48.070 group mean square error. 27:48.360 --> 27:57.540 And what we get out of this is the mean absolute error of one six one zero, the mean square error of 27:57.540 --> 28:04.080 around six fifty eight thousand, then root mean square error of two five, stubbled six. 28:04.290 --> 28:10.680 And if you talk about the accuracy, it is actually eighty six percent, which is of decent accuracy 28:10.980 --> 28:14.850 based on the model and the simplicity of the model which we have taken here. 28:16.860 --> 28:23.910 So this is one implementation of this morning, another implementation, what you can do is you can 28:23.910 --> 28:32.250 simply take the description column and convert them into the idea vectors, as it will provide a lot 28:32.250 --> 28:38.830 more information about Neka, although the information might be present in the different columns. 28:38.850 --> 28:45.840 So that is something which you need to be careful of because then it can introduce a little bias in 28:45.840 --> 28:46.400 the data. 28:46.860 --> 28:50.010 So you can try that as.